();
146 | header.put("Content-Type", sConfig.audioFormat);
147 |
148 | if(sConfig.isAuthNeeded) {
149 | if (this.tokenProvider != null) {
150 | header.put("X-Watson-Authorization-Token", this.tokenProvider.getToken());
151 | Log.d(TAG, "ws connecting with token based authentication");
152 | } else {
153 | String auth = "Basic " + Base64.encodeBytes((this.username + ":" + this.password).getBytes(Charset.forName("UTF-8")));
154 | header.put("Authorization", auth);
155 | Log.d(TAG, "ws connecting with Basic Authentication");
156 | }
157 | }
158 |
159 | if (sConfig.learningOptOut) {
160 | header.put("X-Watson-Learning-OptOut", "true");
161 | Log.d(TAG, "ws setting X-Watson-Learning-OptOut");
162 | }
163 |
164 | String wsURL = getHostURL().toString() + "/v1/recognize" + (this.model != null ? ("?model=" + this.model) : "");
165 |
166 | uploader = new WebSocketUploader(wsURL, header, sConfig);
167 | uploader.setDelegate(this.delegate);
168 | this.startRecording();
169 | } catch (URISyntaxException e) {
170 | e.printStackTrace();
171 | }
172 | }
173 |
174 | /**
175 | * Stop audio recording
176 | */
177 | public void stopRecording(){
178 | if(audioCaptureThread != null)
179 | audioCaptureThread.end();
180 | }
181 |
182 | /**
183 | * Stop recognition
184 | */
185 | public void stopRecognition() {
186 | this.stopRecording();
187 | if(uploader != null) {
188 | uploader.stop();
189 | uploader.close();
190 | }
191 | }
192 |
193 | /**
194 | * Build authentication header
195 | * @param httpGet
196 | */
197 | private void buildAuthenticationHeader(HttpGet httpGet) {
198 | // use token based authentication if possible, otherwise Basic Authentication will be used
199 | if (this.tokenProvider != null) {
200 | Log.d(TAG, "using token based authentication");
201 | httpGet.setHeader("X-Watson-Authorization-Token",this.tokenProvider.getToken());
202 | } else {
203 | Log.d(TAG, "using basic authentication");
204 | httpGet.setHeader(BasicScheme.authenticate(new UsernamePasswordCredentials(this.username, this.password), "UTF-8",false));
205 | }
206 | }
207 |
208 | /**
209 | * Get the list of models for the speech to text service
210 | * @return
211 | */
212 | public JSONObject getModels() {
213 | JSONObject object = null;
214 |
215 | try {
216 | Log.d(TAG, "starting getModels");
217 | HttpClient httpClient = new DefaultHttpClient();
218 | String strHTTPURL = this.hostURL.toString().replace("wss","https").replace("ws", "http");
219 | HttpGet httpGet = new HttpGet(strHTTPURL+"/v1/models");
220 | this.buildAuthenticationHeader(httpGet);
221 | httpGet.setHeader("accept","application/json");
222 | HttpResponse executed = httpClient.execute(httpGet);
223 | InputStream is=executed.getEntity().getContent();
224 |
225 | // get the JSON object containing the models from the InputStream
226 | BufferedReader streamReader = new BufferedReader(new InputStreamReader(is, "UTF-8"));
227 | StringBuilder responseStrBuilder = new StringBuilder();
228 | String inputStr;
229 | while ((inputStr = streamReader.readLine()) != null)
230 | responseStrBuilder.append(inputStr);
231 | Log.d(TAG, "response: " + responseStrBuilder.toString());
232 | object = new JSONObject(responseStrBuilder.toString());
233 | Log.d(TAG, object.toString());
234 | } catch (IOException | JSONException e) {
235 | e.printStackTrace();
236 | }
237 | return object;
238 | }
239 |
240 | /**
241 | * Get information about the model
242 | * @param strModel
243 | * @return
244 | */
245 | public JSONObject getModelInfo(String strModel) {
246 | JSONObject object = null;
247 |
248 | try {
249 | HttpClient httpClient = new DefaultHttpClient();
250 | String strHTTPURL = this.hostURL.toString().replace("wss", "https").replace("ws", "http");
251 | HttpGet httpGet = new HttpGet(strHTTPURL+"/v1/models/en-US_NarrowbandModel");
252 | this.buildAuthenticationHeader(httpGet);
253 | httpGet.setHeader("accept","application/json");
254 | HttpResponse executed = httpClient.execute(httpGet);
255 | InputStream is=executed.getEntity().getContent();
256 |
257 | // get the JSON object containing the models from the InputStream
258 | BufferedReader streamReader = new BufferedReader(new InputStreamReader(is, "UTF-8"));
259 | StringBuilder responseStrBuilder = new StringBuilder();
260 | String inputStr;
261 | while ((inputStr = streamReader.readLine()) != null)
262 | responseStrBuilder.append(inputStr);
263 | object = new JSONObject(responseStrBuilder.toString());
264 | Log.d(TAG, object.toString());
265 |
266 | } catch (IOException | JSONException e) {
267 | e.printStackTrace();
268 | }
269 | return object;
270 | }
271 |
272 | /**
273 | * @return the hostURL
274 | */
275 | public URI getHostURL() {
276 | return hostURL;
277 | }
278 | /**
279 | * @param hostURL the hostURL to set
280 | */
281 | public void setHostURL(URI hostURL) {
282 | this.hostURL = hostURL;
283 | }
284 | /**
285 | * @return the delegate
286 | */
287 | public ISpeechDelegate getDelegate() {
288 | return delegate;
289 | }
290 | /**
291 | * @param val the delegate to set
292 | */
293 | public void setDelegate(ISpeechDelegate val) {
294 | this.delegate = val;
295 | }
296 | /**
297 | * Set API credentials
298 | * @param username
299 | */
300 | public void setCredentials(String username, String password) {
301 | this.username = username;
302 | this.password = password;
303 | }
304 | /**
305 | * Set token provider (for token based authentication)
306 | */
307 | public void setTokenProvider(TokenProvider tokenProvider) { this.tokenProvider = tokenProvider; }
308 | /**
309 | * Set STT model
310 | */
311 | public void setModel(String model) {
312 | this.model = model;
313 | }
314 | }
315 |
316 |
--------------------------------------------------------------------------------
/speech-android-wrapper/src/main/java/com/ibm/watson/developer_cloud/android/speech_to_text/v1/WatsonSDK.java:
--------------------------------------------------------------------------------
1 | /**
2 | * © Copyright IBM Corporation 2015
3 | *
4 | * Licensed under the Apache License, Version 2.0 (the "License");
5 | * you may not use this file except in compliance with the License.
6 | * You may obtain a copy of the License at
7 | *
8 | * http://www.apache.org/licenses/LICENSE-2.0
9 | *
10 | * Unless required by applicable law or agreed to in writing, software
11 | * distributed under the License is distributed on an "AS IS" BASIS,
12 | * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13 | * See the License for the specific language governing permissions and
14 | * limitations under the License.
15 | **/
16 |
17 | package com.ibm.watson.developer_cloud.android.speech_to_text.v1;
18 |
19 | import android.app.Application;
20 |
21 | public class WatsonSDK extends Application {
22 |
23 | }
24 |
--------------------------------------------------------------------------------
/speech-android-wrapper/src/main/java/com/ibm/watson/developer_cloud/android/speech_to_text/v1/audio/AudioCaptureThread.java:
--------------------------------------------------------------------------------
1 | /**
2 | * © Copyright IBM Corporation 2015
3 | *
4 | * Licensed under the Apache License, Version 2.0 (the "License");
5 | * you may not use this file except in compliance with the License.
6 | * You may obtain a copy of the License at
7 | *
8 | * http://www.apache.org/licenses/LICENSE-2.0
9 | *
10 | * Unless required by applicable law or agreed to in writing, software
11 | * distributed under the License is distributed on an "AS IS" BASIS,
12 | * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13 | * See the License for the specific language governing permissions and
14 | * limitations under the License.
15 | **/
16 |
17 | package com.ibm.watson.developer_cloud.android.speech_to_text.v1.audio;
18 |
19 | import java.lang.Thread;
20 | import android.util.Log;
21 | import android.media.AudioRecord;
22 | import android.media.*;
23 | import android.media.MediaRecorder.AudioSource;
24 |
25 | import java.nio.ByteBuffer;
26 | import java.nio.ByteOrder;
27 |
28 | /**
29 | * @author Daniel Bolanos dbolano@us.ibm.com
30 | * description: this thread captures audio from the phone's microphone, whenever the buffer
31 | *
32 | */
33 | public class AudioCaptureThread extends Thread {
34 |
35 | private static final String TAG = "AudioCaptureThread";
36 | private boolean mStop = false;
37 | private boolean mStopped = false;
38 | private int mSamplingRate = -1;
39 | private IAudioConsumer mIAudioConsumer = null;
40 |
41 | // the thread receives high priority because it needs to do real time audio capture
42 | // THREAD_PRIORITY_URGENT_AUDIO = "Standard priority of the most important audio threads"
43 | public AudioCaptureThread(int iSamplingRate, IAudioConsumer IAudioConsumer) {
44 | android.os.Process.setThreadPriority(android.os.Process.THREAD_PRIORITY_URGENT_AUDIO);
45 | mSamplingRate = iSamplingRate;
46 | mIAudioConsumer = IAudioConsumer;
47 | }
48 |
49 | // once the thread is started it runs nonstop until it is stopped from the outside
50 | @Override
51 | public void run() {
52 | AudioRecord recorder = null;
53 |
54 | try {
55 | int iN = Math.max(mSamplingRate/2,AudioRecord.getMinBufferSize(mSamplingRate,AudioFormat.CHANNEL_IN_MONO,AudioFormat.ENCODING_PCM_16BIT));
56 | short[] buffer = new short[iN]; // ASR latency depends on the length of this buffer, a short buffer is good for latency
57 | // because the ASR will process the speech sooner, however it will introduce some network overhead because each packet comes
58 | // with a fixed amount of protocol-data), also I have noticed that some servers cannot handle too many small packages
59 |
60 | // initialize the recorder (buffer size will be at least 1/4th of a second)
61 | recorder = new AudioRecord(AudioSource.MIC, mSamplingRate, AudioFormat.CHANNEL_IN_MONO, AudioFormat.ENCODING_PCM_16BIT, iN);
62 | recorder.startRecording();
63 | Log.d(TAG, "recording started!");
64 | while(!mStop) {
65 |
66 | int r = recorder.read(buffer,0,buffer.length);
67 | long v = 0;
68 | for (int i = 0; i < r; i++) {
69 | v += buffer[i] * buffer[i];
70 | }
71 | double amplitude = v / (double) r;
72 | double volume = 0;
73 | if(amplitude > 0)
74 | volume = 10 * Math.log10(amplitude);
75 | mIAudioConsumer.onAmplitude(amplitude, volume);
76 |
77 | // convert to an array of bytes and send it to the server
78 | ByteBuffer bufferBytes = ByteBuffer.allocate(r*2);
79 | bufferBytes.order(ByteOrder.LITTLE_ENDIAN);
80 | bufferBytes.asShortBuffer().put(buffer,0,r);
81 | byte[] bytes = bufferBytes.array();
82 | int length = bytes.length;
83 | mIAudioConsumer.consume(bytes);
84 | }
85 | }
86 | catch(Throwable x) {
87 | Log.e(TAG, "Error reading voice audio", x);
88 | }
89 | // release resources
90 | finally {
91 | if (recorder != null) {
92 | recorder.stop();
93 | recorder.release();
94 | }
95 | mStopped = true;
96 | Log.d(TAG, "recording stopped!");
97 | }
98 | }
99 |
100 | // this function is intended to be called from outside the thread in order to stop the thread
101 | public void end() {
102 | mStop = true;
103 | // waiting loop, it waits until the thread actually finishes
104 | while(!mStopped) {
105 | try {
106 | Thread.sleep(10);
107 | } catch (InterruptedException e) {
108 | e.printStackTrace();
109 | }
110 | };
111 | }
112 | }
--------------------------------------------------------------------------------
/speech-android-wrapper/src/main/java/com/ibm/watson/developer_cloud/android/speech_to_text/v1/audio/AudioFileWriter.java:
--------------------------------------------------------------------------------
1 | /******************************************************************************
2 | * *
3 | * Copyright (c) 1999-2004 Wimba S.A., All Rights Reserved. *
4 | * *
5 | * COPYRIGHT: *
6 | * This software is the property of Wimba S.A. *
7 | * This software is redistributed under the Xiph.org variant of *
8 | * the BSD license. *
9 | * Redistribution and use in source and binary forms, with or without *
10 | * modification, are permitted provided that the following conditions *
11 | * are met: *
12 | * - Redistributions of source code must retain the above copyright *
13 | * notice, this list of conditions and the following disclaimer. *
14 | * - Redistributions in binary form must reproduce the above copyright *
15 | * notice, this list of conditions and the following disclaimer in the *
16 | * documentation and/or other materials provided with the distribution. *
17 | * - Neither the name of Wimba, the Xiph.org Foundation nor the names of *
18 | * its contributors may be used to endorse or promote products derived *
19 | * from this software without specific prior written permission. *
20 | * *
21 | * WARRANTIES: *
22 | * This software is made available by the authors in the hope *
23 | * that it will be useful, but without any warranty. *
24 | * Wimba S.A. is not liable for any consequence related to the *
25 | * use of the provided software. *
26 | * *
27 | * Class: AudioFileWriter.java *
28 | * *
29 | * Author: Marc GIMPEL *
30 | * *
31 | * Date: 6th January 2004 *
32 | * *
33 | ******************************************************************************/
34 |
35 | /* $Id: AudioFileWriter.java,v 1.2 2004/10/21 16:21:57 mgimpel Exp $ */
36 |
37 | package com.ibm.watson.developer_cloud.android.speech_to_text.v1.audio;
38 |
39 | import java.io.DataOutput;
40 | import java.io.File;
41 | import java.io.IOException;
42 | import java.io.OutputStream;
43 |
44 | /**
45 | * Abstract Class that defines an Audio File Writer.
46 | *
47 | * @author Marc Gimpel, Wimba S.A. (mgimpel@horizonwimba.com)
48 | * @version $Revision: 1.2 $
49 | */
50 | public abstract class AudioFileWriter{
51 | /**
52 | * Closes the output file.
53 | * @exception IOException if there was an exception closing the Audio Writer.
54 | */
55 | public abstract void close() throws IOException;
56 |
57 | /**
58 | * Open the output file.
59 | * @param file - file to open.
60 | * @exception IOException if there was an exception opening the Audio Writer.
61 | */
62 | public abstract void open(File file) throws IOException;
63 |
64 | /**
65 | * Open the output file.
66 | * @param filename - file to open.
67 | * @exception IOException if there was an exception opening the Audio Writer.
68 | */
69 | public abstract void open(String filename) throws IOException;
70 |
71 | /**
72 | * Writes the header pages that start the Ogg Opus file.
73 | * Prepares file for data to be written.
74 | * @param comment description to be included in the header.
75 | * @exception IOException
76 | */
77 | public abstract void writeHeader(String comment) throws IOException;
78 |
79 | /**
80 | * Writes a packet of audio.
81 | * @param data audio data
82 | * @param offset the offset from which to start reading the data.
83 | * @param len the length of data to read.
84 | * @exception IOException
85 | */
86 | public abstract void writePacket(byte[] data, int offset, int len) throws IOException;
87 |
88 | /**
89 | * Writes an Ogg Page Header to the given byte array.
90 | * @param buf the buffer to write to.
91 | * @param offset the from which to start writing.
92 | * @param headerType the header type flag
93 | * (0=normal, 2=bos: beginning of stream, 4=eos: end of stream).
94 | * @param granulepos the absolute granule position.
95 | * @param streamSerialNumber
96 | * @param pageCount
97 | * @param packetCount
98 | * @param packetSizes
99 | * @return the amount of data written to the buffer.
100 | */
101 | public static int writeOggPageHeader(byte[] buf, int offset, int headerType,
102 | long granulepos, int streamSerialNumber,
103 | int pageCount, int packetCount,
104 | byte[] packetSizes)
105 | {
106 | writeString(buf, offset, "OggS"); // 0 - 3: capture_pattern
107 | buf[offset+4] = 0; // 4: stream_structure_version
108 | buf[offset+5] = (byte) headerType; // 5: header_type_flag
109 | writeLong(buf, offset+6, granulepos); // 6 - 13: absolute granule position
110 | writeInt(buf, offset+14, streamSerialNumber); // 14 - 17: stream serial number
111 | writeInt(buf, offset+18, pageCount); // 18 - 21: page sequence no
112 | writeInt(buf, offset+22, 0); // 22 - 25: page checksum
113 | buf[offset+26] = (byte) packetCount; // 26: page_segments
114 | System.arraycopy(packetSizes, 0, // 27 - x: segment_table
115 | buf, offset + 27, packetCount);
116 | return packetCount+27;
117 | }
118 |
119 | /**
120 | * Builds and returns an Ogg Page Header.
121 | * @param headerType the header type flag
122 | * (0=normal, 2=bos: beginning of stream, 4=eos: end of stream).
123 | * @param granulepos the absolute granule position.
124 | * @param streamSerialNumber
125 | * @param pageCount
126 | * @param packetCount
127 | * @param packetSizes
128 | * @return an Ogg Page Header.
129 | */
130 | public static byte[] buildOggPageHeader(int headerType, long granulepos,
131 | int streamSerialNumber, int pageCount,
132 | int packetCount, byte[] packetSizes)
133 | {
134 | byte[] data = new byte[packetCount+27];
135 | writeOggPageHeader(data, 0, headerType, granulepos, streamSerialNumber,
136 | pageCount, packetCount, packetSizes);
137 | return data;
138 | }
139 |
140 | /**
141 | * Builds a Opus Header.
142 | *
143 | * @param sampleRate
144 | * @return Opus Header data
145 | */
146 | public static byte[] buildOpusHeader(int sampleRate) {
147 | byte[] data = new byte[19];
148 | writeOpusHeader(data, 0, sampleRate);
149 | return data;
150 | }
151 |
152 | /**
153 | * Write Opus header
154 | *
155 | * @param buf
156 | * @param offset
157 | * @param sampleRate
158 | *
159 | * @link https://tools.ietf.org/html/draft-ietf-codec-oggopus-08#section-5.1
160 | */
161 | public static void writeOpusHeader(byte[] buf, int offset, int sampleRate) {
162 | // Magic Signature
163 | writeString(buf, offset, "OpusHead");
164 | buf[offset + 8] = 1; // Version, MUST The version number MUST always be '1' for this version of the encapsulation specification.
165 | buf[offset + 9] = 1; // Output Channel Count
166 | writeShort(buf, offset + 10, 0); // Pre-skip
167 | writeInt(buf, offset + 12, sampleRate); // Input Sample Rate (Hz)
168 | writeShort(buf, offset + 16, 0); // Output Gain (Q7.8 in dB), +/- 128 dB
169 | buf[offset + 18] = 0; // Mapping Family (For channel mapping family 0, this value defaults to C-1 (i.e., 0 for mono and 1 for stereo), and is not coded.)
170 | }
171 |
172 | /**
173 | * Writes a Opus Comment to the given byte array.
174 | *
175 | * @param buf
176 | * the buffer to write to.
177 | * @param offset
178 | * the from which to start writing.
179 | * @param comment
180 | * the comment.
181 | * @return the amount of data written to the buffer.
182 | */
183 | public static void writeOpusComment(byte[] buf, int offset, String comment) {
184 | // Magic Signature
185 | writeString(buf, offset, "OpusTags");
186 | String vendorString = "IBM";
187 | writeInt(buf, offset + 8, vendorString.length()); // Vendor String Length
188 | writeString(buf, offset + 12, vendorString); // Vendor String
189 | writeInt(buf, offset + 20, 1); // User Comment List Length
190 | writeInt(buf, offset + 24, comment.length()); // User Comment #0 String Length
191 | writeString(buf, offset + 28, comment); // User Comment #0 String
192 | }
193 |
194 | /**
195 | * Builds and returns a Opus Comment.
196 | *
197 | * @param comment
198 | * the comment.
199 | * @return a Opus Comment.
200 | */
201 | public static byte[] buildOpusComment(String comment) {
202 | byte[] data = new byte[28+comment.length()];
203 | writeOpusComment(data, 0, comment);
204 | return data;
205 | }
206 |
207 | /**
208 | * Writes a Little-endian short.
209 | * @param out the data output to write to.
210 | * @param v value to write.
211 | * @exception IOException
212 | */
213 | public static void writeShort(DataOutput out, short v)
214 | throws IOException
215 | {
216 | out.writeByte((0xff & v));
217 | out.writeByte((0xff & (v >>> 8)));
218 | }
219 |
220 | /**
221 | * Writes a Little-endian int.
222 | * @param out the data output to write to.
223 | * @param v value to write.
224 | * @exception IOException
225 | */
226 | public static void writeInt(DataOutput out, int v)
227 | throws IOException
228 | {
229 | out.writeByte(0xff & v);
230 | out.writeByte(0xff & (v >>> 8));
231 | out.writeByte(0xff & (v >>> 16));
232 | out.writeByte(0xff & (v >>> 24));
233 | }
234 |
235 | /**
236 | * Writes a Little-endian short.
237 | * @param os - the output stream to write to.
238 | * @param v - the value to write.
239 | * @exception IOException
240 | */
241 | public static void writeShort(OutputStream os, short v)
242 | throws IOException
243 | {
244 | os.write((0xff & v));
245 | os.write((0xff & (v >>> 8)));
246 | }
247 |
248 | /**
249 | * Writes a Little-endian int.
250 | * @param os - the output stream to write to.
251 | * @param v - the value to write.
252 | * @exception IOException
253 | */
254 | public static void writeInt(OutputStream os, int v)
255 | throws IOException
256 | {
257 | os.write(0xff & v);
258 | os.write(0xff & (v >>> 8));
259 | os.write(0xff & (v >>> 16));
260 | os.write(0xff & (v >>> 24));
261 | }
262 |
263 | /**
264 | * Writes a Little-endian long.
265 | * @param os - the output stream to write to.
266 | * @param v - the value to write.
267 | * @exception IOException
268 | */
269 | public static void writeLong(OutputStream os, long v)
270 | throws IOException
271 | {
272 | os.write((int)(0xff & v));
273 | os.write((int)(0xff & (v >>> 8)));
274 | os.write((int)(0xff & (v >>> 16)));
275 | os.write((int)(0xff & (v >>> 24)));
276 | os.write((int)(0xff & (v >>> 32)));
277 | os.write((int)(0xff & (v >>> 40)));
278 | os.write((int)(0xff & (v >>> 48)));
279 | os.write((int) (0xff & (v >>> 56)));
280 | }
281 |
282 | /**
283 | * Writes a Little-endian short.
284 | * @param data the array into which the data should be written.
285 | * @param offset the offset from which to start writing in the array.
286 | * @param v the value to write.
287 | */
288 | public static void writeShort(byte[] data, int offset, int v)
289 | {
290 | data[offset] = (byte) (0xff & v);
291 | data[offset+1] = (byte) (0xff & (v >>> 8));
292 | }
293 |
294 | /**
295 | * Writes a Little-endian int.
296 | * @param data the array into which the data should be written.
297 | * @param offset the offset from which to start writing in the array.
298 | * @param v the value to write.
299 | */
300 | public static void writeInt(byte[] data, int offset, int v)
301 | {
302 | data[offset] = (byte) (0xff & v);
303 | data[offset+1] = (byte) (0xff & (v >>> 8));
304 | data[offset+2] = (byte) (0xff & (v >>> 16));
305 | data[offset+3] = (byte) (0xff & (v >>> 24));
306 | }
307 |
308 | /**
309 | * Writes a Little-endian long.
310 | * @param data the array into which the data should be written.
311 | * @param offset the offset from which to start writing in the array.
312 | * @param v the value to write.
313 | */
314 | public static void writeLong(byte[] data, int offset, long v)
315 | {
316 | data[offset] = (byte) (0xff & v);
317 | data[offset+1] = (byte) (0xff & (v >>> 8));
318 | data[offset+2] = (byte) (0xff & (v >>> 16));
319 | data[offset+3] = (byte) (0xff & (v >>> 24));
320 | data[offset+4] = (byte) (0xff & (v >>> 32));
321 | data[offset+5] = (byte) (0xff & (v >>> 40));
322 | data[offset+6] = (byte) (0xff & (v >>> 48));
323 | data[offset+7] = (byte) (0xff & (v >>> 56));
324 | }
325 |
326 | /**
327 | * Writes a String.
328 | * @param data the array into which the data should be written.
329 | * @param offset the offset from which to start writing in the array.
330 | * @param v the value to write.
331 | */
332 | public static void writeString(byte[] data, int offset, String v)
333 | {
334 | byte[] str = v.getBytes();
335 | System.arraycopy(str, 0, data, offset, str.length);
336 | }
337 | }
338 |
--------------------------------------------------------------------------------
/speech-android-wrapper/src/main/java/com/ibm/watson/developer_cloud/android/speech_to_text/v1/audio/IAudioConsumer.java:
--------------------------------------------------------------------------------
1 | /**
2 | * © Copyright IBM Corporation 2015
3 | *
4 | * Licensed under the Apache License, Version 2.0 (the "License");
5 | * you may not use this file except in compliance with the License.
6 | * You may obtain a copy of the License at
7 | *
8 | * http://www.apache.org/licenses/LICENSE-2.0
9 | *
10 | * Unless required by applicable law or agreed to in writing, software
11 | * distributed under the License is distributed on an "AS IS" BASIS,
12 | * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13 | * See the License for the specific language governing permissions and
14 | * limitations under the License.
15 | **/
16 |
17 | package com.ibm.watson.developer_cloud.android.speech_to_text.v1.audio;
18 |
19 | /**
20 | * Created by daniel bolanos on 8/24/15.
21 | */
22 | public interface IAudioConsumer {
23 | // Use PROPRIETARY notice if class contains a main() method, otherwise use COPYRIGHT notice.
24 | public static final String COPYRIGHT_NOTICE = "(c) Copyright IBM Corp. 2015";
25 | // function that consumes the audio data
26 | public void consume(byte [] data);
27 | public void onAmplitude(double amplitude, double volume);
28 | }
29 |
--------------------------------------------------------------------------------
/speech-android-wrapper/src/main/java/com/ibm/watson/developer_cloud/android/speech_to_text/v1/audio/IChunkUploader.java:
--------------------------------------------------------------------------------
1 | /**
2 | * © Copyright IBM Corporation 2015
3 | *
4 | * Licensed under the Apache License, Version 2.0 (the "License");
5 | * you may not use this file except in compliance with the License.
6 | * You may obtain a copy of the License at
7 | *
8 | * http://www.apache.org/licenses/LICENSE-2.0
9 | *
10 | * Unless required by applicable law or agreed to in writing, software
11 | * distributed under the License is distributed on an "AS IS" BASIS,
12 | * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13 | * See the License for the specific language governing permissions and
14 | * limitations under the License.
15 | **/
16 |
17 | package com.ibm.watson.developer_cloud.android.speech_to_text.v1.audio;
18 |
19 | import com.ibm.watson.developer_cloud.android.speech_to_text.v1.ISpeechDelegate;
20 |
21 | /**
22 | * The uploader interface.
23 | */
24 | public interface IChunkUploader {
25 | // Use PROPRIETARY notice if class contains a main() method, otherwise use COPYRIGHT notice.
26 | public static final String COPYRIGHT_NOTICE = "(c) Copyright IBM Corp. 2015";
27 | /**
28 | * On has data.
29 | *
30 | * @param buffer the buffer
31 | */
32 | int onHasData(byte[] buffer);
33 | /**
34 | * Checks if uploader has been prepared.
35 | *
36 | * @return true, if uploader is prepared
37 | */
38 | boolean isUploadPrepared();
39 | /**
40 | * Upload data
41 | * @param data
42 | */
43 | void upload(byte[] data);
44 | /**
45 | * Stop uploading
46 | */
47 | void stop();
48 | /**
49 | * Start thread to construct an upload http connection to back end server.
50 | */
51 | void prepare();
52 |
53 | /**
54 | * Set Delegate
55 | *
56 | * @param delegate
57 | */
58 | void setDelegate(ISpeechDelegate delegate);
59 | /**
60 | * Close connection.
61 | */
62 | void close();
63 | }
64 |
--------------------------------------------------------------------------------
/speech-android-wrapper/src/main/java/com/ibm/watson/developer_cloud/android/speech_to_text/v1/audio/ISpeechEncoder.java:
--------------------------------------------------------------------------------
1 | /**
2 | * © Copyright IBM Corporation 2015
3 | *
4 | * Licensed under the Apache License, Version 2.0 (the "License");
5 | * you may not use this file except in compliance with the License.
6 | * You may obtain a copy of the License at
7 | *
8 | * http://www.apache.org/licenses/LICENSE-2.0
9 | *
10 | * Unless required by applicable law or agreed to in writing, software
11 | * distributed under the License is distributed on an "AS IS" BASIS,
12 | * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13 | * See the License for the specific language governing permissions and
14 | * limitations under the License.
15 | **/
16 |
17 | package com.ibm.watson.developer_cloud.android.speech_to_text.v1.audio;
18 |
19 | import java.io.IOException;
20 |
21 | /**
22 | * Encoder interface.
23 | */
24 | public interface ISpeechEncoder {
25 | // Use PROPRIETARY notice if class contains a main() method, otherwise use COPYRIGHT notice.
26 | public static final String COPYRIGHT_NOTICE = "(c) Copyright IBM Corp. 2015";
27 | /**
28 | * Init encoder with the WebSocket client
29 | * @param uploader
30 | * @throws IOException
31 | */
32 | void initEncoderWithUploader(IChunkUploader uploader) throws IOException;
33 | /**
34 | * In compression mode, encode audio data (to SPX) before write to ouput stream.
35 | * In non-compression mode, write directly raw audio data to ouput stream.
36 | *
37 | * @param b audio data will be written
38 | * @throws IOException Signals that an I/O exception has occurred.
39 | */
40 | int encodeAndWrite(byte[] b) throws IOException ;
41 | /**
42 | * Get compression audio time in compression mode.
43 | * @return the time for compression audio.
44 | */
45 | void onStart();
46 | /**
47 | * Close output stream.
48 | */
49 | void close();
50 | }
51 |
--------------------------------------------------------------------------------
/speech-android-wrapper/src/main/java/com/ibm/watson/developer_cloud/android/speech_to_text/v1/audio/OggCrc.java:
--------------------------------------------------------------------------------
1 | /******************************************************************************
2 | * *
3 | * Copyright (c) 1999-2003 Wimba S.A., All Rights Reserved. *
4 | * *
5 | * COPYRIGHT: *
6 | * This software is the property of Wimba S.A. *
7 | * This software is redistributed under the Xiph.org variant of *
8 | * the BSD license. *
9 | * Redistribution and use in source and binary forms, with or without *
10 | * modification, are permitted provided that the following conditions *
11 | * are met: *
12 | * - Redistributions of source code must retain the above copyright *
13 | * notice, this list of conditions and the following disclaimer. *
14 | * - Redistributions in binary form must reproduce the above copyright *
15 | * notice, this list of conditions and the following disclaimer in the *
16 | * documentation and/or other materials provided with the distribution. *
17 | * - Neither the name of Wimba, the Xiph.org Foundation nor the names of *
18 | * its contributors may be used to endorse or promote products derived *
19 | * from this software without specific prior written permission. *
20 | * *
21 | * WARRANTIES: *
22 | * This software is made available by the authors in the hope *
23 | * that it will be useful, but without any warranty. *
24 | * Wimba S.A. is not liable for any consequence related to the *
25 | * use of the provided software. *
26 | * *
27 | * Class: OggCrc.java *
28 | * *
29 | * Author: Marc GIMPEL *
30 | * Based on code by: Ross WILLIAMS *
31 | * *
32 | * Date: 20th April 2003 *
33 | * *
34 | ******************************************************************************/
35 |
36 | /* $Id: OggCrc.java,v 1.2 2004/10/21 16:21:57 mgimpel Exp $ */
37 |
38 | /********************************************************************
39 | * *
40 | * THIS FILE IS PART OF THE OggVorbis SOFTWARE CODEC SOURCE CODE. *
41 | * USE, DISTRIBUTION AND REPRODUCTION OF THIS LIBRARY SOURCE IS *
42 | * GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE *
43 | * IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING. *
44 | * *
45 | * THE OggVorbis SOURCE CODE IS (C) COPYRIGHT 1994-2002 *
46 | * by the Xiph.Org Foundation http://www.xiph.org/ *
47 | * *
48 | ********************************************************************
49 |
50 | function: code raw [Vorbis] packets into framed OggSquish stream and
51 | decode Ogg streams back into raw packets
52 | last mod: $Id: OggCrc.java,v 1.2 2004/10/21 16:21:57 mgimpel Exp $
53 |
54 | note: The CRC code is directly derived from public domain code by
55 | Ross Williams (ross@guest.adelaide.edu.au). See docs/framing.html
56 | for details.
57 |
58 | ********************************************************************/
59 |
60 | package com.ibm.watson.developer_cloud.android.speech_to_text.v1.audio;
61 |
62 | /**
63 | * Calculates the CRC checksum for Ogg packets.
64 | *
65 | * Ogg uses the same generator polynomial as ethernet, although with an
66 | * unreflected alg and an init/final of 0, not 0xffffffff.
67 | *
68 | * @author Jim Lawrence, helloNetwork.com
69 | * @author Marc Gimpel, Wimba S.A. (mgimpel@horizonwimba.com)
70 | * @version $Revision: 1.2 $
71 | */
72 | public class OggCrc
73 | {
74 | // TODO - implement java.util.zip.Checksum
75 |
76 | /**
77 | * CRC checksum lookup table
78 | */
79 | private static int[] crc_lookup;
80 |
81 | static {
82 | crc_lookup = new int[256];
83 | for (int i=0; i>>24)&0xff)^(data[offset]&0xff)];
120 | }
121 | return crc;
122 | }
123 | }
124 |
--------------------------------------------------------------------------------
/speech-android-wrapper/src/main/java/com/ibm/watson/developer_cloud/android/speech_to_text/v1/audio/OggOpusEnc.java:
--------------------------------------------------------------------------------
1 | /**
2 | * © Copyright IBM Corporation 2015
3 | *
4 | * Licensed under the Apache License, Version 2.0 (the "License");
5 | * you may not use this file except in compliance with the License.
6 | * You may obtain a copy of the License at
7 | *
8 | * http://www.apache.org/licenses/LICENSE-2.0
9 | *
10 | * Unless required by applicable law or agreed to in writing, software
11 | * distributed under the License is distributed on an "AS IS" BASIS,
12 | * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13 | * See the License for the specific language governing permissions and
14 | * limitations under the License.
15 | **/
16 |
17 | package com.ibm.watson.developer_cloud.android.speech_to_text.v1.audio;
18 |
19 | import com.ibm.watson.developer_cloud.android.speech_to_text.v1.dto.SpeechConfiguration;
20 | import com.ibm.watson.developer_cloud.android.speech_to_text.v1.opus.JNAOpus;
21 | import com.ibm.watson.developer_cloud.android.speech_to_text.v1.opus.OpusWriter;
22 | import com.sun.jna.ptr.PointerByReference;
23 |
24 | import java.io.ByteArrayInputStream;
25 | import java.io.IOException;
26 | import java.nio.ByteBuffer;
27 | import java.nio.IntBuffer;
28 | import java.nio.ShortBuffer;
29 |
30 | /**
31 | * Ogg Opus Encoder
32 | */
33 | public class OggOpusEnc extends OpusWriter implements ISpeechEncoder {
34 | // Use PROPRIETARY notice if class contains a main() method, otherwise use COPYRIGHT notice.
35 | public static final String COPYRIGHT_NOTICE = "(c) Copyright IBM Corp. 2015";
36 | /** Data writer */
37 | private OpusWriter writer = null;
38 | /** Opus encoder reference */
39 | private PointerByReference opusEncoder;
40 | /**
41 | * Constructor
42 | */
43 | public OggOpusEnc() {}
44 | /**
45 | * For WebSocketClient
46 | * @param uploader
47 | * @throws IOException
48 | */
49 | public void initEncoderWithUploader(IChunkUploader uploader) throws IOException{
50 | writer = new OpusWriter(uploader);
51 |
52 | IntBuffer error = IntBuffer.allocate(4);
53 | this.opusEncoder = JNAOpus.INSTANCE.opus_encoder_create(
54 | SpeechConfiguration.SAMPLE_RATE,
55 | SpeechConfiguration.AUDIO_CHANNELS,
56 | JNAOpus.OPUS_APPLICATION_VOIP,
57 | error);
58 | }
59 | /**
60 | * When the encode begins
61 | */
62 | @Override
63 | public void onStart() {
64 | writer.writeHeader("encoder=Lavc56.20.100 libopus");
65 | }
66 | /**
67 | * Encode raw audio data into Opus format then call OpusWriter to write the Ogg packet
68 | *
69 | * @param rawAudio
70 | * @return
71 | * @throws IOException
72 | */
73 | public int encodeAndWrite(byte[] rawAudio) throws IOException {
74 | int uploadedAudioSize = 0;
75 | ByteArrayInputStream ios = new ByteArrayInputStream(rawAudio);
76 |
77 | byte[] data = new byte[SpeechConfiguration.FRAME_SIZE*2];
78 | int bufferSize, read;
79 |
80 | while((read = ios.read(data)) > 0){
81 | bufferSize = read;
82 | byte[] pcmBuffer = new byte[read];
83 | System.arraycopy(data, 0, pcmBuffer, 0, read);
84 |
85 | ShortBuffer shortBuffer = ShortBuffer.allocate(bufferSize);
86 | for (int i = 0; i < read; i += 2) {
87 | int b1 = pcmBuffer[i] & 0xff;
88 | int b2 = pcmBuffer[i+1] << 8;
89 | shortBuffer.put((short) (b1 | b2));
90 | }
91 | shortBuffer.flip();
92 | ByteBuffer opusBuffer = ByteBuffer.allocate(bufferSize);
93 |
94 | int opus_encoded = JNAOpus.INSTANCE.opus_encode(this.opusEncoder, shortBuffer, SpeechConfiguration.FRAME_SIZE, opusBuffer, bufferSize);
95 |
96 | opusBuffer.position(opus_encoded);
97 | opusBuffer.flip();
98 |
99 | byte[] opusData = new byte[opusBuffer.remaining()];
100 | opusBuffer.get(opusData, 0, opusData.length);
101 |
102 | if (opus_encoded > 0) {
103 | uploadedAudioSize += opusData.length;
104 | writer.writePacket(opusData, 0, opusData.length);
105 | }
106 | }
107 |
108 | ios.close();
109 | return uploadedAudioSize;
110 | }
111 | /**
112 | * Close writer
113 | */
114 | public void close() {
115 | try {
116 | writer.close();
117 | JNAOpus.INSTANCE.opus_encoder_destroy(this.opusEncoder);
118 | } catch (IOException e) {
119 | e.printStackTrace();
120 | }
121 | }
122 | }
123 |
--------------------------------------------------------------------------------
/speech-android-wrapper/src/main/java/com/ibm/watson/developer_cloud/android/speech_to_text/v1/audio/PcmWaveWriter.java:
--------------------------------------------------------------------------------
1 | /******************************************************************************
2 | * *
3 | * Copyright (c) 1999-2003 Wimba S.A., All Rights Reserved. *
4 | * *
5 | * COPYRIGHT: *
6 | * This software is the property of Wimba S.A. *
7 | * This software is redistributed under the Xiph.org variant of *
8 | * the BSD license. *
9 | * Redistribution and use in source and binary forms, with or without *
10 | * modification, are permitted provided that the following conditions *
11 | * are met: *
12 | * - Redistributions of source code must retain the above copyright *
13 | * notice, this list of conditions and the following disclaimer. *
14 | * - Redistributions in binary form must reproduce the above copyright *
15 | * notice, this list of conditions and the following disclaimer in the *
16 | * documentation and/or other materials provided with the distribution. *
17 | * - Neither the name of Wimba, the Xiph.org Foundation nor the names of *
18 | * its contributors may be used to endorse or promote products derived *
19 | * from this software without specific prior written permission. *
20 | * *
21 | * WARRANTIES: *
22 | * This software is made available by the authors in the hope *
23 | * that it will be useful, but without any warranty. *
24 | * Wimba S.A. is not liable for any consequence related to the *
25 | * use of the provided software. *
26 | * *
27 | * Class: PcmWaveWriter.java *
28 | * *
29 | * Author: James LAWRENCE *
30 | * Modified by: Marc GIMPEL *
31 | * *
32 | * Date: March 2003 *
33 | * *
34 | ******************************************************************************/
35 |
36 | /* $Id: PcmWaveWriter.java,v 1.2 2004/10/21 16:21:57 mgimpel Exp $ */
37 |
38 | package com.ibm.watson.developer_cloud.android.speech_to_text.v1.audio;
39 |
40 | import android.util.Log;
41 |
42 | import java.io.ByteArrayOutputStream;
43 | import java.io.File;
44 | import java.io.IOException;
45 | import java.io.RandomAccessFile;
46 |
47 | /**
48 | * Writes basic PCM wave files from binary audio data.
49 | *
50 | * Here's an example that writes 2 seconds of silence
51 | *
52 | * PcmWaveWriter s_wsw = new PcmWaveWriter(2, 44100);
53 | * byte[] silence = new byte[16*2*44100];
54 | * wsw.Open("C:\\out.wav");
55 | * wsw.WriteHeader();
56 | * wsw.WriteData(silence, 0, silence.length);
57 | * wsw.WriteData(silence, 0, silence.length);
58 | * wsw.Close();
59 | *
60 | *
61 | * @author Jim Lawrence, helloNetwork.com
62 | * @author Marc Gimpel, Wimba S.A. (mgimpel@horizonwimba.com)
63 | * @version $Revision: 1.2 $
64 | */
65 | public class PcmWaveWriter extends AudioFileWriter {
66 |
67 | private static final String TAG = "PcmWaveWriter";
68 |
69 | /** Wave type code of PCM */
70 | public static final short WAVE_FORMAT_PCM = (short) 0x01;
71 | private RandomAccessFile raf;
72 | /** Defines the sampling rate of the audio input. */
73 | private int sampleRate;
74 | /** Defines the number of channels of the audio input (1=mono, 2=stereo). */
75 | private int channels;
76 |
77 | /**
78 | * Constructor.
79 | */
80 | public PcmWaveWriter() {}
81 |
82 | /**
83 | * Constructor.
84 | * @param sampleRate the number of samples per second.
85 | * @param channels the number of audio channels (1=mono, 2=stereo, ...).
86 | */
87 | public PcmWaveWriter(final int sampleRate, final int channels)
88 | {
89 | this();
90 | this.channels = channels;
91 | this.sampleRate = sampleRate;
92 | }
93 |
94 | /**
95 | * Closes the output file.
96 | * MUST be called to have a correct stream.
97 | * @exception IOException if there was an exception closing the Audio Writer.
98 | */
99 | public void close()
100 | throws IOException
101 | {
102 | raf.close();
103 | }
104 |
105 | /**
106 | * Open the output file.
107 | * @param file - file to open.
108 | * @exception IOException if there was an exception opening the Audio Writer.
109 | */
110 | public void open(final File file)
111 | throws IOException
112 | {
113 | file.delete();
114 | raf = new RandomAccessFile(file, "rw");
115 | }
116 |
117 | /**
118 | * Open the output file.
119 | * @param filename filename to open.
120 | * @exception IOException if there was an exception opening the Audio Writer.
121 | */
122 | public void open(final String filename)
123 | throws IOException
124 | {
125 | open(new File(filename));
126 | }
127 |
128 | /**
129 | * Writes the initial data chunks that start the wave file.
130 | * Prepares file for data samples to written.
131 | * @param comment ignored by the WAV header.
132 | * @exception IOException
133 | */
134 | public void writeHeader(final String comment)
135 | throws IOException
136 | {
137 | /* writes the RIFF chunk indicating wave format */
138 | byte[] chkid = "RIFF".getBytes();
139 | raf.write(chkid, 0, chkid.length);
140 | writeInt(raf, 0); /* total length must be blank */
141 | chkid = "WAVE".getBytes();
142 | raf.write(chkid, 0, chkid.length);
143 |
144 | /* format subchunk: of size 16 */
145 | chkid = "fmt ".getBytes();
146 | raf.write(chkid, 0, chkid.length);
147 |
148 | writeInt(raf, 16); // Size of format chunk
149 | writeShort(raf, WAVE_FORMAT_PCM); // Format tag: PCM
150 | writeShort(raf, (short) channels); // Number of channels
151 | writeInt(raf, sampleRate); // Sampling frequency
152 | writeInt(raf, sampleRate * channels * 2); // Average bytes per second
153 | writeShort(raf, (short) (channels * 2)); // Blocksize of data
154 | writeShort(raf, (short) 16); // Bit per sample
155 | /* write the start of data chunk */
156 | chkid = "data".getBytes();
157 | raf.write(chkid, 0, chkid.length);
158 | writeInt(raf, 0);
159 | }
160 |
161 | /**
162 | * Saves PCM data to WAV file
163 | * @param pcmdata
164 | * : byte array containing the PCM data
165 | * @param srate
166 | * : Sample rate
167 | * @param channel
168 | * : no. of channels
169 | * @param format
170 | * : PCM format (16 bit)
171 | * @throws IOException
172 | */
173 | public byte[] saveWav(byte[] pcmdata, int srate, int channel, int format) {
174 |
175 | byte[] header = new byte[44];
176 | byte[] data = pcmdata;
177 |
178 | long totalDataLen = data.length + 36;
179 | long bitrate = srate * channel * format;
180 |
181 | header[0] = 'R';
182 | header[1] = 'I';
183 | header[2] = 'F';
184 | header[3] = 'F';
185 | header[4] = (byte) (totalDataLen & 0xff);
186 | header[5] = (byte) ((totalDataLen >> 8) & 0xff);
187 | header[6] = (byte) ((totalDataLen >> 16) & 0xff);
188 | header[7] = (byte) ((totalDataLen >> 24) & 0xff);
189 | header[8] = 'W';
190 | header[9] = 'A';
191 | header[10] = 'V';
192 | header[11] = 'E';
193 | header[12] = 'f';
194 | header[13] = 'm';
195 | header[14] = 't';
196 | header[15] = ' ';
197 | header[16] = (byte) format;
198 | header[17] = 0;
199 | header[18] = 0;
200 | header[19] = 0;
201 | header[20] = 1;
202 | header[21] = 0;
203 | header[22] = (byte) channel;
204 | header[23] = 0;
205 | header[24] = (byte) (srate & 0xff);
206 | header[25] = (byte) ((srate >> 8) & 0xff);
207 | header[26] = (byte) ((srate >> 16) & 0xff);
208 | header[27] = (byte) ((srate >> 24) & 0xff);
209 | header[28] = (byte) ((bitrate / 8) & 0xff);
210 | header[29] = (byte) (((bitrate / 8) >> 8) & 0xff);
211 | header[30] = (byte) (((bitrate / 8) >> 16) & 0xff);
212 | header[31] = (byte) (((bitrate / 8) >> 24) & 0xff);
213 | header[32] = (byte) ((channel * format) / 8);
214 | header[33] = 0;
215 | header[34] = 16;
216 | header[35] = 0;
217 | header[36] = 'd';
218 | header[37] = 'a';
219 | header[38] = 't';
220 | header[39] = 'a';
221 | header[40] = (byte) (data.length & 0xff);
222 | header[41] = (byte) ((data.length >> 8) & 0xff);
223 | header[42] = (byte) ((data.length >> 16) & 0xff);
224 | header[43] = (byte) ((data.length >> 24) & 0xff);
225 |
226 | ByteArrayOutputStream outputStream = new ByteArrayOutputStream( );
227 | try {
228 | outputStream.write(header);
229 | outputStream.write(data);
230 |
231 | } catch (IOException e) {
232 | Log.e(TAG, "Error writing data to wav buffer");
233 | e.printStackTrace();
234 | }
235 | return outputStream.toByteArray();
236 | }
237 |
238 | /**
239 | * Saves PCM data to WAV file
240 | * @param pcmdata
241 | * : byte array containing the PCM data
242 | * @param srate
243 | * : Sample rate
244 | * @param channel
245 | * : no. of channels
246 | * @param format
247 | * : PCM format (16 bit)
248 | * @throws IOException
249 | */
250 | public void saveWavFile(byte[] pcmdata, int srate, int channel, int format) {
251 |
252 | byte[] header = new byte[44];
253 | byte[] data = pcmdata;
254 |
255 | long totalDataLen = data.length + 36;
256 | long bitrate = srate * channel * format;
257 |
258 | header[0] = 'R';
259 | header[1] = 'I';
260 | header[2] = 'F';
261 | header[3] = 'F';
262 | header[4] = (byte) (totalDataLen & 0xff);
263 | header[5] = (byte) ((totalDataLen >> 8) & 0xff);
264 | header[6] = (byte) ((totalDataLen >> 16) & 0xff);
265 | header[7] = (byte) ((totalDataLen >> 24) & 0xff);
266 | header[8] = 'W';
267 | header[9] = 'A';
268 | header[10] = 'V';
269 | header[11] = 'E';
270 | header[12] = 'f';
271 | header[13] = 'm';
272 | header[14] = 't';
273 | header[15] = ' ';
274 | header[16] = (byte) format;
275 | header[17] = 0;
276 | header[18] = 0;
277 | header[19] = 0;
278 | header[20] = 1;
279 | header[21] = 0;
280 | header[22] = (byte) channel;
281 | header[23] = 0;
282 | header[24] = (byte) (srate & 0xff);
283 | header[25] = (byte) ((srate >> 8) & 0xff);
284 | header[26] = (byte) ((srate >> 16) & 0xff);
285 | header[27] = (byte) ((srate >> 24) & 0xff);
286 | header[28] = (byte) ((bitrate / 8) & 0xff);
287 | header[29] = (byte) (((bitrate / 8) >> 8) & 0xff);
288 | header[30] = (byte) (((bitrate / 8) >> 16) & 0xff);
289 | header[31] = (byte) (((bitrate / 8) >> 24) & 0xff);
290 | header[32] = (byte) ((channel * format) / 8);
291 | header[33] = 0;
292 | header[34] = 16;
293 | header[35] = 0;
294 | header[36] = 'd';
295 | header[37] = 'a';
296 | header[38] = 't';
297 | header[39] = 'a';
298 | header[40] = (byte) (data.length & 0xff);
299 | header[41] = (byte) ((data.length >> 8) & 0xff);
300 | header[42] = (byte) ((data.length >> 16) & 0xff);
301 | header[43] = (byte) ((data.length >> 24) & 0xff);
302 |
303 | try {
304 | raf.write(header, 0, 44);
305 | raf.write(data);
306 | raf.close();
307 | } catch (IOException e) {
308 | Log.e(TAG, "Error writing data to wav file");
309 | e.printStackTrace();
310 | }
311 |
312 | Log.e(TAG, "wrote Wav File");
313 | }
314 |
315 | /**
316 | * Writes a packet of audio.
317 | * @param data audio data
318 | * @param offset the offset from which to start reading the data.
319 | * @param len the length of data to read.
320 | * @exception IOException
321 | */
322 | public void writePacket(final byte[] data,
323 | final int offset,
324 | final int len)
325 | throws IOException
326 | {
327 | raf.write(data, offset, len);
328 | }
329 | }
--------------------------------------------------------------------------------
/speech-android-wrapper/src/main/java/com/ibm/watson/developer_cloud/android/speech_to_text/v1/audio/RawEnc.java:
--------------------------------------------------------------------------------
1 | /**
2 | * © Copyright IBM Corporation 2015
3 | *
4 | * Licensed under the Apache License, Version 2.0 (the "License");
5 | * you may not use this file except in compliance with the License.
6 | * You may obtain a copy of the License at
7 | *
8 | * http://www.apache.org/licenses/LICENSE-2.0
9 | *
10 | * Unless required by applicable law or agreed to in writing, software
11 | * distributed under the License is distributed on an "AS IS" BASIS,
12 | * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13 | * See the License for the specific language governing permissions and
14 | * limitations under the License.
15 | **/
16 |
17 | package com.ibm.watson.developer_cloud.android.speech_to_text.v1.audio;
18 |
19 | import java.io.IOException;
20 |
21 | /**
22 | * Raw data encoder
23 | */
24 | public class RawEnc implements ISpeechEncoder {
25 | // Use PROPRIETARY notice if class contains a main() method, otherwise use COPYRIGHT notice.
26 | public static final String COPYRIGHT_NOTICE = "(c) Copyright IBM Corp. 2015";
27 | /** Data writer */
28 | private RawWriter writer = null;
29 | /**
30 | * Constructor.
31 | */
32 | public RawEnc() {}
33 | /**
34 | * For WebSocketClient
35 | * @param uploader
36 | * @throws java.io.IOException
37 | */
38 | public void initEncoderWithUploader(IChunkUploader uploader) throws IOException{
39 | this.writer = new RawWriter(uploader);
40 | }
41 | /**
42 | * On encode begin
43 | */
44 | @Override
45 | public void onStart() {}
46 | /* (non-Javadoc)
47 | * @see com.ibm.watson.developer_cloud.android.speech_to_text.v1.audio.SpeechEncoder#encodeAndWrite(byte[])
48 | */
49 | @Override
50 | public int encodeAndWrite(byte[] b) throws IOException {
51 | writer.writePacket(b, 0, b.length);
52 | return b.length;
53 | }
54 | /* (non-Javadoc)
55 | * @see com.ibm.watson.developer_cloud.android.speech_to_text.v1.audio.SpeechEncoder#close()
56 | */
57 | public void close() {
58 | if(this.writer != null){
59 | try {
60 | this.writer.close();
61 | } catch (IOException e) {
62 | e.printStackTrace();
63 | }
64 | }
65 | }
66 | }
67 |
--------------------------------------------------------------------------------
/speech-android-wrapper/src/main/java/com/ibm/watson/developer_cloud/android/speech_to_text/v1/audio/RawWriter.java:
--------------------------------------------------------------------------------
1 | /**
2 | * Copyright IBM Corporation 2015
3 | *
4 | * Licensed under the Apache License, Version 2.0 (the "License");
5 | * you may not use this file except in compliance with the License.
6 | * You may obtain a copy of the License at
7 | *
8 | * http://www.apache.org/licenses/LICENSE-2.0
9 | *
10 | * Unless required by applicable law or agreed to in writing, software
11 | * distributed under the License is distributed on an "AS IS" BASIS,
12 | * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13 | * See the License for the specific language governing permissions and
14 | * limitations under the License.
15 | **/
16 |
17 | package com.ibm.watson.developer_cloud.android.speech_to_text.v1.audio;
18 |
19 | import java.io.File;
20 | import java.io.IOException;
21 |
22 |
23 | /**
24 | * @author Viney
25 | *
26 | */
27 | public class RawWriter extends AudioFileWriter{
28 | // Use PROPRIETARY notice if class contains a main() method, otherwise use COPYRIGHT notice.
29 | public static final String COPYRIGHT_NOTICE = "(c) Copyright IBM Corp. 2015";
30 | private IChunkUploader uploader;
31 |
32 | /**
33 | * Construct with WebSocketClient
34 | */
35 | public RawWriter(IChunkUploader uploader) {
36 | this.uploader = uploader;
37 | }
38 |
39 | @Override
40 | public void close() throws IOException {
41 | this.uploader.stop();
42 | }
43 |
44 | @Override
45 | public void open(File file) throws IOException {}
46 |
47 |
48 | @Override
49 | public void open(String filename) throws IOException {}
50 |
51 |
52 | @Override
53 | public void writeHeader(String comment) throws IOException {}
54 |
55 | @Override
56 | public void writePacket(byte[] data, int offset, int len) throws IOException {
57 | this.uploader.upload(data);
58 | }
59 | }
60 |
--------------------------------------------------------------------------------
/speech-android-wrapper/src/main/java/com/ibm/watson/developer_cloud/android/speech_to_text/v1/audio/WebSocketUploader.java:
--------------------------------------------------------------------------------
1 | /**
2 | * © Copyright IBM Corporation 2015
3 | *
4 | * Licensed under the Apache License, Version 2.0 (the "License");
5 | * you may not use this file except in compliance with the License.
6 | * You may obtain a copy of the License at
7 | *
8 | * http://www.apache.org/licenses/LICENSE-2.0
9 | *
10 | * Unless required by applicable law or agreed to in writing, software
11 | * distributed under the License is distributed on an "AS IS" BASIS,
12 | * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13 | * See the License for the specific language governing permissions and
14 | * limitations under the License.
15 | **/
16 |
17 |
18 | package com.ibm.watson.developer_cloud.android.speech_to_text.v1.audio;
19 |
20 | import java.io.IOException;
21 | import java.net.URI;
22 | import java.net.URISyntaxException;
23 | import java.nio.channels.NotYetConnectedException;
24 | import java.security.KeyManagementException;
25 | import java.security.NoSuchAlgorithmException;
26 | import java.security.cert.CertificateException;
27 | import java.security.cert.X509Certificate;
28 | import java.util.Map;
29 |
30 | import javax.net.ssl.SSLContext;
31 | import javax.net.ssl.SSLSocketFactory;
32 | import javax.net.ssl.TrustManager;
33 | import javax.net.ssl.X509TrustManager;
34 |
35 | import org.java_websocket.client.WebSocketClient;
36 | import org.java_websocket.drafts.Draft_17;
37 | import org.java_websocket.handshake.ServerHandshake;
38 | import org.json.JSONException;
39 | import org.json.JSONObject;
40 |
41 | import android.util.Log;
42 |
43 | import com.ibm.watson.developer_cloud.android.speech_to_text.v1.dto.SpeechConfiguration;
44 | import com.ibm.watson.developer_cloud.android.speech_to_text.v1.ISpeechDelegate;
45 |
46 | public class WebSocketUploader extends WebSocketClient implements IChunkUploader {
47 | // Use PROPRIETARY notice if class contains a main() method, otherwise use COPYRIGHT notice.
48 | public static final String COPYRIGHT_NOTICE = "(c) Copyright IBM Corp. 2015";
49 | private static final String TAG = WebSocketUploader.class.getName();
50 |
51 | private ISpeechEncoder encoder = null;
52 | private Thread initStreamToServerThread;
53 |
54 | private boolean uploadPrepared = false;
55 |
56 | /** STT delegate */
57 | private ISpeechDelegate delegate = null;
58 | /** Recorder delegate */
59 | private SpeechConfiguration sConfig = null;
60 |
61 | /**
62 | * Create an uploader which supports streaming.
63 | *
64 | * @param serverURL LMC server, delivery to back end server
65 | * @throws URISyntaxException
66 | */
67 | public WebSocketUploader(String serverURL, Map header, SpeechConfiguration config) throws URISyntaxException {
68 | super(new URI(serverURL), new Draft_17(), header, config.connectionTimeout);
69 | Log.d(TAG, "New WebSocketUploader: " + serverURL);
70 | Log.d(TAG, serverURL);
71 | this.sConfig = config;
72 |
73 | if(sConfig.audioFormat.equals(SpeechConfiguration.AUDIO_FORMAT_DEFAULT)) {
74 | this.encoder = new RawEnc();
75 | }
76 | else if(sConfig.audioFormat.equals(SpeechConfiguration.AUDIO_FORMAT_OGGOPUS)){
77 | this.encoder = new OggOpusEnc();
78 | }
79 |
80 | if(serverURL.toLowerCase().startsWith("wss") || serverURL.toLowerCase().startsWith("https"))
81 | this.sConfig.isSSL = true;
82 | else this.sConfig.isSSL = false;
83 | }
84 | /**
85 | * Trust server
86 | *
87 | * @throws KeyManagementException
88 | * @throws NoSuchAlgorithmException
89 | */
90 | private void trustServer() throws KeyManagementException, NoSuchAlgorithmException, IOException {
91 | // Create a trust manager that does not validate certificate chains
92 | TrustManager[] certs = new TrustManager[]{ new X509TrustManager() {
93 | public java.security.cert.X509Certificate[] getAcceptedIssuers() {
94 | return new java.security.cert.X509Certificate[]{};
95 | }
96 | public void checkClientTrusted(X509Certificate[] chain, String authType) throws CertificateException {}
97 | public void checkServerTrusted(X509Certificate[] chain, String authType) throws CertificateException {}
98 | }};
99 | SSLContext sslContext = null;
100 | sslContext = SSLContext.getInstance("TLS");
101 | sslContext.init(null, certs, new java.security.SecureRandom());
102 | SSLSocketFactory factory = sslContext.getSocketFactory();
103 | this.setSocket(factory.createSocket());
104 | }
105 | /**
106 | * 1. Initialize WebSocket connection
107 | * 2. Init an encoder and writer
108 | *
109 | * @throws Exception
110 | */
111 | private void initStreamAudioToServer() throws Exception {
112 | Log.d(TAG, "Connecting...");
113 | //lifted up for initializing writer, using isRunning to control the flow
114 | this.encoder.initEncoderWithUploader(this);
115 |
116 | if(this.sConfig.isSSL)
117 | this.trustServer();
118 |
119 | boolean rc;
120 | rc = this.connectBlocking();
121 |
122 | if (rc) {
123 | Log.d(TAG, "Connected");
124 | this.sendSpeechHeader();
125 | }
126 | else{
127 | Log.e(TAG, "Connection failed!");
128 | this.uploadPrepared = false;
129 | throw new Exception("Connection failed!");
130 | }
131 |
132 | }
133 |
134 | @Override
135 | public int onHasData(byte[] buffer) {
136 | int uploadedAudioSize = 0;
137 | // NOW, WE HAVE STATUS OF UPLOAD PREPARING, UPLOAD PREPARING OK
138 | if (this.isUploadPrepared()) {
139 | try {
140 | uploadedAudioSize = encoder.encodeAndWrite(buffer);
141 | Log.d(TAG, "onHasData: " + uploadedAudioSize + " " + buffer.length);
142 | // TODO: Capturing data
143 | } catch (IOException e) {
144 | e.printStackTrace();
145 | }
146 | }
147 | else {
148 | try {
149 | Log.w(TAG, "waiting for establishing the connection");
150 | initStreamToServerThread.join();
151 | }
152 | catch (InterruptedException e) {
153 | e.printStackTrace();
154 | }
155 | }
156 | return uploadedAudioSize;
157 | }
158 |
159 | @Override
160 | public boolean isUploadPrepared() {
161 | return this.uploadPrepared;
162 | }
163 |
164 | public void stopUploaderPrepareThread() {
165 | if (initStreamToServerThread != null) {
166 | initStreamToServerThread.interrupt();
167 | }
168 | }
169 | /**
170 | * Prepare connection
171 | */
172 | @Override
173 | public void prepare() {
174 | this.uploadPrepared = false;
175 | initStreamToServerThread = new Thread() {
176 | public void run() {
177 | try {
178 | try {
179 | initStreamAudioToServer();
180 | Log.d(TAG, "WebSocket Connection established");
181 | } catch (IOException e1) {
182 | Log.e(TAG, "IOException: " + e1.getMessage());
183 | throw e1;
184 | } catch (InterruptedException e1) {
185 | Log.e(TAG, "InterruptedException:" + e1.getMessage());
186 | throw e1;
187 | } catch (Exception e1) {
188 | Log.e(TAG, "Exception: " + e1.getMessage());
189 | throw e1;
190 | }
191 | } catch (Exception e) {
192 | e.printStackTrace();
193 | Log.e(TAG, "Connection failed: " + (e == null ? "null exception" : e.getMessage()));
194 | uploadPrepared = false;
195 | close();
196 | }
197 | };
198 | };
199 | initStreamToServerThread.setName("initStreamToServerThread");
200 | initStreamToServerThread.start();
201 | }
202 |
203 | /**
204 | * Write string into socket
205 | *
206 | * @param message
207 | */
208 | public void upload(String message){
209 | try{
210 | this.send(message);
211 | }
212 | catch(NotYetConnectedException ex){
213 | Log.e(TAG, ex.getLocalizedMessage());
214 | }
215 | }
216 |
217 | /**
218 | * Write data into socket
219 | *
220 | * @param data
221 | */
222 | public void upload(byte[] data){
223 | try{
224 | this.send(data);
225 | }
226 | catch(NotYetConnectedException ex){
227 | Log.e(TAG, ex.getLocalizedMessage());
228 | }
229 | }
230 |
231 | /**
232 | * Stop by sending out zero byte of data
233 | */
234 | public void stop(){
235 | byte[] stopData = new byte[0];
236 | this.upload(stopData);
237 | }
238 |
239 | @Override
240 | public void close() {
241 | Log.d(TAG, "closing the websocket");
242 | super.close();
243 | }
244 |
245 | @Override
246 | public void onClose(int code, String reason, boolean remote) {
247 | Log.d(TAG, "WebSocket closed");
248 | this.uploadPrepared = false;
249 |
250 | Log.d(TAG, "### Code: " + code + " reason: " + reason + " remote: " + remote);
251 | if (delegate != null){
252 | delegate.onClose(code, reason, remote);
253 | }
254 | }
255 |
256 | @Override
257 | public void onError(Exception ex) {
258 | Log.e(TAG, "WebSocket error");
259 | String errorMessage = "";
260 | if(ex != null)
261 | errorMessage = ex.getMessage();
262 | // Send the error message to the delegate
263 | this.uploadPrepared = false;
264 | //this.sendMessage(ISpeechDelegate.ERROR);
265 | if (delegate != null){
266 | delegate.onError(errorMessage);
267 | }
268 | }
269 |
270 | @Override
271 | public void onMessage(String message) {
272 |
273 | Log.d(TAG + "onMessage", message);
274 | if (delegate != null){
275 | delegate.onMessage(message);
276 | }
277 | }
278 |
279 | @Override
280 | public void onOpen(ServerHandshake arg0) {
281 | Log.d(TAG, "WS connection opened successfully");
282 | this.uploadPrepared = true;
283 | if (delegate != null){
284 | delegate.onOpen();
285 | }
286 | }
287 |
288 | private void sendSpeechHeader() {
289 |
290 | JSONObject obj = new JSONObject();
291 | try {
292 | obj.put("action", "start");
293 | obj.put("content-type", this.sConfig.audioFormat);
294 | obj.put("interim_results", this.sConfig.returnInterimResults);
295 | obj.put("continuous", this.sConfig.continuous);
296 | obj.put("inactivity_timeout", this.sConfig.inactivityTimeout);
297 |
298 | if (this.sConfig.maxAlternatives > 1) {
299 | obj.put("max_alternatives", this.sConfig.maxAlternatives);
300 | }
301 |
302 | if (!(Float.isNaN(this.sConfig.wordAlternativesThreshold))) {
303 | obj.put("word_alternatives_threshold", this.sConfig.wordAlternativesThreshold);
304 | }
305 | } catch (JSONException e) {
306 | e.printStackTrace();
307 | }
308 | String startHeader = obj.toString();
309 | this.upload(startHeader);
310 | this.encoder.onStart();
311 | Log.d(TAG, "Sending init message: " + startHeader);
312 | }
313 |
314 | /**
315 | * Set delegate
316 | *
317 | * @param delegate
318 | */
319 | public void setDelegate(ISpeechDelegate delegate) {
320 | this.delegate = delegate;
321 | }
322 | }
323 |
--------------------------------------------------------------------------------
/speech-android-wrapper/src/main/java/com/ibm/watson/developer_cloud/android/speech_to_text/v1/dto/SpeechConfiguration.java:
--------------------------------------------------------------------------------
1 | /**
2 | * © Copyright IBM Corporation 2015
3 | *
4 | * Licensed under the Apache License, Version 2.0 (the "License");
5 | * you may not use this file except in compliance with the License.
6 | * You may obtain a copy of the License at
7 | *
8 | * http://www.apache.org/licenses/LICENSE-2.0
9 | *
10 | * Unless required by applicable law or agreed to in writing, software
11 | * distributed under the License is distributed on an "AS IS" BASIS,
12 | * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13 | * See the License for the specific language governing permissions and
14 | * limitations under the License.
15 | **/
16 |
17 |
18 | package com.ibm.watson.developer_cloud.android.speech_to_text.v1.dto;
19 |
20 | /**
21 | * Created by mihui on 9/2/15.
22 | */
23 | public class SpeechConfiguration {
24 |
25 | // PCM format
26 | public static final String AUDIO_FORMAT_DEFAULT = "audio/l16;rate=16000";
27 | // OggOpus format
28 | public static final String AUDIO_FORMAT_OGGOPUS = "audio/ogg;codecs=opus";
29 | // Audio channels
30 | public static final int AUDIO_CHANNELS = 1;
31 | // Frame size
32 | public static final int FRAME_SIZE = 160;
33 | // Sample rate
34 | public static final int SAMPLE_RATE = 16000;
35 | // Timeout
36 | public int inactivityTimeout = 600;
37 | // Data format
38 | public String audioFormat = AUDIO_FORMAT_DEFAULT;
39 | // Authentication flag
40 | public boolean isAuthNeeded = true;
41 | // SSL flag, this would be detected automatically
42 | public boolean isSSL = true;
43 | // Default timeout duration for a connection
44 | public int connectionTimeout = 30000;
45 | // Interim results are intermediate hypotheses of a transcription that are likely to change before the service returns the final result. Interim results are useful for real-time transcription, for example, to obtain alternative transcriptions for a dictation application.
46 | public boolean returnInterimResults = true;
47 | // An integer value that tells the service to return the n-best alternative hypotheses
48 | public int maxAlternatives = 1;
49 | // Reports hypotheses for acoustically similar alternatives for words of the input audio
50 | public float wordAlternativesThreshold = Float.NaN;
51 | // Indicates whether multiple final results that represent consecutive phrases separated by long pauses are returned
52 | public boolean continuous = true;
53 | // Indicates whether to opt out of data collection for requests sent over the connection
54 | public boolean learningOptOut = false;
55 |
56 | /**
57 | * Instantiate default configuration
58 | */
59 | public SpeechConfiguration(){}
60 |
61 | /**
62 | * Constructing configuration by parameters
63 | *
64 | * @param audioFormat
65 | */
66 | public SpeechConfiguration(String audioFormat){
67 | this.audioFormat = audioFormat;
68 | }
69 |
70 | /**
71 | * Constructing configuration by parameters
72 | *
73 | * @param audioFormat
74 | * @param isAuthNeeded
75 | */
76 | public SpeechConfiguration(String audioFormat, boolean isAuthNeeded){
77 | this.audioFormat = audioFormat;
78 | this.isAuthNeeded = isAuthNeeded;
79 | }
80 | }
81 |
--------------------------------------------------------------------------------
/speech-android-wrapper/src/main/java/com/ibm/watson/developer_cloud/android/speech_to_text/v1/opus/OggOpus.java:
--------------------------------------------------------------------------------
1 | /**
2 | * © Copyright IBM Corporation 2015
3 | *
4 | * Licensed under the Apache License, Version 2.0 (the "License");
5 | * you may not use this file except in compliance with the License.
6 | * You may obtain a copy of the License at
7 | *
8 | * http://www.apache.org/licenses/LICENSE-2.0
9 | *
10 | * Unless required by applicable law or agreed to in writing, software
11 | * distributed under the License is distributed on an "AS IS" BASIS,
12 | * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13 | * See the License for the specific language governing permissions and
14 | * limitations under the License.
15 | **/
16 |
17 | package com.ibm.watson.developer_cloud.android.speech_to_text.v1.opus;
18 |
19 | public class OggOpus {
20 | public static native void initAudio();
21 | public static native void startRecorder( int sample_rate);
22 | public static native void stopRecorder();
23 | public static native int encode( String s, int sample_rate );
24 | public static native int decode( String s, String o, int sample_rate );
25 | public static native float volume();
26 | static {
27 | System.loadLibrary("oggopus");
28 | }
29 | }
30 |
--------------------------------------------------------------------------------
/speech-android-wrapper/src/main/java/com/ibm/watson/developer_cloud/android/speech_to_text/v1/opus/OpusWriter.java:
--------------------------------------------------------------------------------
1 | /**
2 | * © Copyright IBM Corporation 2015
3 | *
4 | * Licensed under the Apache License, Version 2.0 (the "License");
5 | * you may not use this file except in compliance with the License.
6 | * You may obtain a copy of the License at
7 | *
8 | * http://www.apache.org/licenses/LICENSE-2.0
9 | *
10 | * Unless required by applicable law or agreed to in writing, software
11 | * distributed under the License is distributed on an "AS IS" BASIS,
12 | * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13 | * See the License for the specific language governing permissions and
14 | * limitations under the License.
15 | **/
16 |
17 | package com.ibm.watson.developer_cloud.android.speech_to_text.v1.opus;
18 |
19 | import java.io.File;
20 | import java.io.IOException;
21 | import java.util.Random;
22 |
23 | import com.ibm.watson.developer_cloud.android.speech_to_text.v1.audio.AudioFileWriter;
24 | import com.ibm.watson.developer_cloud.android.speech_to_text.v1.audio.IChunkUploader;
25 | import com.ibm.watson.developer_cloud.android.speech_to_text.v1.audio.OggCrc;
26 |
27 | import com.ibm.watson.developer_cloud.android.speech_to_text.v1.dto.SpeechConfiguration;
28 |
29 | public class OpusWriter extends AudioFileWriter {
30 | private String TAG = this.getClass().getSimpleName();
31 | private IChunkUploader uploader;
32 | /** Number of packets in an Ogg page (must be less than 255) */
33 | public static final int PACKETS_PER_OGG_PAGE = 50;
34 | /** Defines the sampling rate of the audio input. */
35 | protected int sampleRate;
36 | /** Ogg Stream Serial Number */
37 | protected int streamSerialNumber;
38 | /** Data buffer */
39 | private byte[] dataBuffer;
40 | /** Pointer within the Data buffer */
41 | private int dataBufferPtr;
42 | /** Header buffer */
43 | private byte[] headerBuffer;
44 | /** Pointer within the Header buffer */
45 | private int headerBufferPtr;
46 | /** Ogg Page count */
47 | protected int pageCount;
48 | /** Opus packet count within an Ogg Page */
49 | private int packetCount;
50 | /**
51 | * Absolute granule position
52 | * (the number of audio samples from beginning of file to end of Ogg Packet).
53 | */
54 | private long granulepos;
55 | /** Frame size */
56 | private int frameSize;
57 |
58 | public OpusWriter(){}
59 |
60 | /**
61 | * Setting up the OggOpus Writer
62 | * @param uploader
63 | */
64 | public OpusWriter(IChunkUploader uploader){
65 | this.uploader = uploader;
66 | if (streamSerialNumber == 0)
67 | streamSerialNumber = new Random().nextInt();
68 | dataBuffer = new byte[65565];
69 | dataBufferPtr = 0;
70 | headerBuffer = new byte[255];
71 | headerBufferPtr = 0;
72 | pageCount = 0;
73 | packetCount = 0;
74 | granulepos = 0;
75 | this.sampleRate = SpeechConfiguration.SAMPLE_RATE;
76 | this.frameSize = SpeechConfiguration.FRAME_SIZE;
77 | }
78 | @Override
79 | public void close() throws IOException {
80 | flush(true);
81 | this.uploader.stop();
82 | }
83 |
84 | @Override
85 | public void open(File file) throws IOException {}
86 |
87 | @Override
88 | public void open(String filename) throws IOException {}
89 |
90 | @Override
91 | public void writeHeader(String comment) {
92 | byte[] header;
93 | byte[] data;
94 | int chkSum;
95 |
96 | /* writes the OGG header page */
97 | header = buildOggPageHeader(2, 0, streamSerialNumber, pageCount++, 1, new byte[] {19});
98 | data = buildOpusHeader(sampleRate);
99 | chkSum = OggCrc.checksum(0, header, 0, header.length);
100 | chkSum = OggCrc.checksum(chkSum, data, 0, data.length);
101 | writeInt(header, 22, chkSum);
102 | this.write(header);
103 | this.write(data);
104 |
105 | /* Writes the OGG comment page */
106 | header = buildOggPageHeader(0, 0, streamSerialNumber, pageCount++, 1, new byte[]{(byte) (comment.length() + 8)});
107 | data = buildOpusComment(comment);
108 | chkSum = OggCrc.checksum(0, header, 0, header.length);
109 | chkSum = OggCrc.checksum(chkSum, data, 0, data.length);
110 | writeInt(header, 22, chkSum);
111 | this.write(header);
112 | this.write(data);
113 | }
114 |
115 | /**
116 | * Write data packet
117 | * @param data audio data
118 | * @param offset the offset from which to start reading the data.
119 | * @param len the length of data to read.
120 | * @throws IOException
121 | */
122 | @Override
123 | public void writePacket(byte[] data, int offset, int len)
124 | throws IOException {
125 | // if nothing to write
126 | if (len <= 0) {
127 | return;
128 | }
129 | if (packetCount > PACKETS_PER_OGG_PAGE) {
130 | flush(false);
131 | }
132 | System.arraycopy(data, offset, dataBuffer, dataBufferPtr, len);
133 | dataBufferPtr += len;
134 | headerBuffer[headerBufferPtr++]=(byte)len;
135 | packetCount++;
136 | granulepos += this.frameSize*2;
137 | }
138 |
139 | /**
140 | * Flush the Ogg page out of the buffers into the file.
141 | * @param eos - end of stream
142 | * @exception IOException
143 | */
144 | protected void flush(final boolean eos) throws IOException{
145 | int chksum;
146 | byte[] header;
147 | /* Writes the OGG header page */
148 | header = buildOggPageHeader((eos ? 4 : 0), granulepos, streamSerialNumber, pageCount++, packetCount, headerBuffer);
149 | chksum = OggCrc.checksum(0, header, 0, header.length);
150 | chksum = OggCrc.checksum(chksum, dataBuffer, 0, dataBufferPtr);
151 | writeInt(header, 22, chksum);
152 |
153 | this.write(header);
154 | this.write(dataBuffer, 0, dataBufferPtr);
155 |
156 | dataBufferPtr = 0;
157 | headerBufferPtr = 0;
158 | packetCount = 0;
159 | }
160 |
161 | /**
162 | * Write data into Socket
163 | * @param data
164 | */
165 | public void write(byte[] data){
166 | this.uploader.upload(data);
167 | }
168 |
169 | /**
170 | * Write data into Socket
171 | * @param data
172 | * @param offset
173 | * @param count
174 | */
175 | public void write(byte[] data, int offset, int count){
176 | byte[] tmp = new byte[count];
177 | System.arraycopy(data, offset, tmp, 0, count);
178 |
179 | this.uploader.upload(tmp);
180 | }
181 | }
182 |
--------------------------------------------------------------------------------
/speech-android-wrapper/src/main/java/com/ibm/watson/developer_cloud/android/text_to_speech/v1/TTSUtility.java:
--------------------------------------------------------------------------------
1 | /**
2 | * © Copyright IBM Corporation 2015
3 | *
4 | * Licensed under the Apache License, Version 2.0 (the "License");
5 | * you may not use this file except in compliance with the License.
6 | * You may obtain a copy of the License at
7 | *
8 | * http://www.apache.org/licenses/LICENSE-2.0
9 | *
10 | * Unless required by applicable law or agreed to in writing, software
11 | * distributed under the License is distributed on an "AS IS" BASIS,
12 | * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13 | * See the License for the specific language governing permissions and
14 | * limitations under the License.
15 | **/
16 |
17 | package com.ibm.watson.developer_cloud.android.text_to_speech.v1;
18 |
19 | import android.app.Application;
20 | import android.media.AudioFormat;
21 | import android.media.AudioManager;
22 | import android.media.AudioTrack;
23 | import android.os.Environment;
24 | import android.util.Log;
25 |
26 | import com.ibm.watson.developer_cloud.android.speech_to_text.v1.opus.OggOpus;
27 |
28 | import org.apache.commons.io.IOUtils;
29 | import org.apache.http.HttpResponse;
30 | import org.apache.http.auth.UsernamePasswordCredentials;
31 | import org.apache.http.client.HttpClient;
32 | import org.apache.http.client.methods.HttpGet;
33 | import org.apache.http.client.utils.URLEncodedUtils;
34 | import org.apache.http.impl.auth.BasicScheme;
35 | import org.apache.http.impl.client.DefaultHttpClient;
36 | import org.apache.http.message.BasicNameValuePair;
37 |
38 | import com.ibm.watson.developer_cloud.android.speech_to_text.v1.audio.PcmWaveWriter;
39 |
40 | import java.io.File;
41 |
42 | import java.io.FileNotFoundException;
43 |
44 | import java.io.IOException;
45 | import java.io.InputStream;
46 |
47 | import java.io.RandomAccessFile;
48 |
49 | import java.util.LinkedList;
50 | import java.util.List;
51 |
52 | public class TTSUtility extends Application {
53 | private static final String TAG = TTSUtility.class.getName();
54 |
55 | public static final String CODEC_WAV = "audio/wav";
56 | public static final int CODEC_WAV_SAMPLE_RATE = 0;
57 |
58 | public static final String CODEC_OPUS = "audio/opus";
59 | public static final int CODEC_OPUS_SAMPLE_RATE = 48000;
60 |
61 | private String username;
62 | private String password;
63 | private String token;
64 | private boolean learningOptOut;
65 | private String voice;
66 | private String content;
67 | private String codec;
68 | private int sampleRate;
69 | private String server;
70 | private AudioTrack audioTrack;
71 |
72 |
73 | public TTSUtility(){
74 | this.codec = CODEC_WAV;
75 | // By default, the sample rate would be detected by the SDK if the value is set to zero
76 | // However, the metadata is not reliable, need to decode at the maximum sample rate
77 | this.sampleRate = 48000;
78 | }
79 |
80 | /**
81 | * Set codec
82 | * @param codec
83 | */
84 | public void setCodec(String codec){
85 | this.codec = codec;
86 | }
87 |
88 | /**
89 | * Stop player
90 | */
91 | private void stopTtsPlayer() {
92 | if (audioTrack != null && audioTrack.getState() != AudioTrack.STATE_UNINITIALIZED ) {
93 | // IMPORTANT: NOT use stop()
94 | // For an immediate stop, use pause(), followed by flush() to discard audio data that hasn't been played back yet.
95 | audioTrack.pause();
96 | audioTrack.flush();
97 | }
98 | }
99 |
100 | /**
101 | * Text to speech
102 | * @param arguments
103 | */
104 | public void synthesize(String[] arguments) {
105 | Log.i(TAG, "Start requesting TTS... ("+this.codec+")");
106 | try {
107 | parseParams(arguments);
108 | } catch (Exception e) {
109 | e.printStackTrace();
110 | }
111 |
112 | if(this.codec == CODEC_WAV){
113 | this.sampleRate = CODEC_WAV_SAMPLE_RATE;
114 | }
115 | else{
116 | this.sampleRate = CODEC_OPUS_SAMPLE_RATE;
117 | }
118 |
119 | TTSThread thread = new TTSThread();
120 | thread.setName("TTSThread");
121 | thread.start();
122 | }
123 |
124 | private void initPlayer(){
125 | stopTtsPlayer();
126 | // IMPORTANT: minimum required buffer size for the successful creation of an AudioTrack instance in streaming mode.
127 | int bufferSize = AudioTrack.getMinBufferSize(sampleRate, AudioFormat.CHANNEL_OUT_MONO, AudioFormat.ENCODING_PCM_16BIT);
128 |
129 | synchronized (this) {
130 | audioTrack = new AudioTrack(AudioManager.STREAM_MUSIC,
131 | sampleRate,
132 | AudioFormat.CHANNEL_OUT_MONO,
133 | AudioFormat.ENCODING_PCM_16BIT,
134 | bufferSize,
135 | AudioTrack.MODE_STREAM);
136 | if (audioTrack != null)
137 | audioTrack.play();
138 | }
139 | }
140 |
141 | private void parseParams(String[] arguments){
142 | int i = 0;
143 | this.server = arguments[i++];
144 | this.username = arguments[i++];
145 | this.password = arguments[i++];
146 | this.voice = arguments[i++];
147 | this.content = arguments[i++];
148 | this.token = arguments[i++];
149 | this.learningOptOut = Boolean.valueOf(arguments[i++]);
150 | }
151 |
152 | /**
153 | * Post text data to the server and get returned audio data
154 | * @param server iTrans server
155 | * @param username
156 | * @param password
157 | * @param content
158 | * @return {@link HttpResponse}
159 | * @throws Exception
160 | */
161 | public static HttpResponse createPost(String server, String username, String password, String token, boolean learningOptOut, String content, String voice, String codec) throws Exception {
162 | String url = server;
163 |
164 | //HTTP GET Client
165 | HttpClient httpClient = new DefaultHttpClient();
166 | //Add params
167 | List params = new LinkedList();
168 | params.add(new BasicNameValuePair("text", content));
169 | params.add(new BasicNameValuePair("voice", voice));
170 | params.add(new BasicNameValuePair("accept", codec));
171 | HttpGet httpGet = new HttpGet(url+"?"+ URLEncodedUtils.format(params, "utf-8"));
172 | // use token based authentication if possible, otherwise Basic Authentication will be used
173 | if (token != null) {
174 | Log.d(TAG, "using token based authentication");
175 | httpGet.setHeader("X-Watson-Authorization-Token",token);
176 | } else {
177 | Log.d(TAG, "using basic authentication");
178 | httpGet.setHeader(BasicScheme.authenticate(new UsernamePasswordCredentials(username, password), "UTF-8", false));
179 | }
180 |
181 | if (learningOptOut) {
182 | Log.d(TAG, "setting X-Watson-Learning-OptOut");
183 | httpGet.setHeader("X-Watson-Learning-Opt-Out", "true");
184 | }
185 | HttpResponse executed = httpClient.execute(httpGet);
186 | return executed;
187 | }
188 |
189 | /**
190 | * Get storage path
191 | * @return
192 | */
193 | private String getBaseDir() {
194 | String baseDir;
195 | if (Environment.getExternalStorageState().equals(Environment.MEDIA_MOUNTED)) {
196 | baseDir = Environment.getExternalStorageDirectory().getAbsolutePath() + "/";
197 | } else {
198 | baseDir = "/data/data/" + getApplicationContext().getPackageName() + "/";
199 | }
200 |
201 | return baseDir;
202 | }
203 |
204 | /**
205 | * Thread to post text data to iTrans server and play returned audio data
206 | * @author chienlk
207 | *
208 | */
209 | public class TTSThread extends Thread {
210 | @Override
211 | public void run() {
212 | android.os.Process.setThreadPriority(android.os.Process.THREAD_PRIORITY_URGENT_AUDIO);
213 |
214 | HttpResponse post;
215 | try {
216 | post = createPost(server, username, password, token, learningOptOut, content, voice, codec);
217 | InputStream is = post.getEntity().getContent();
218 |
219 | byte[] data = null;
220 | if(codec == CODEC_WAV) {
221 | data = analyzeWavData(is);
222 | }
223 | else if(codec == CODEC_OPUS){
224 | data = analyzeOpusData(is);
225 | }
226 | initPlayer();
227 | audioTrack.write(data, 0, data.length);
228 | is.close();
229 |
230 | } catch (Exception e) {
231 | e.printStackTrace();
232 | } finally {
233 | Log.i(TAG, "Stopping audioTrack...");
234 | if (audioTrack != null && audioTrack.getState() != AudioTrack.STATE_UNINITIALIZED) {
235 | audioTrack.release();
236 | }
237 | }
238 | }
239 | }
240 |
241 | private byte[] analyzeOpusData(InputStream is) {
242 | String inFilePath = getBaseDir()+"Watson.opus";
243 | String outFilePath = getBaseDir()+"Watson.pcm";
244 | File inFile = new File(inFilePath);
245 | File outFile = new File(outFilePath);
246 | outFile.deleteOnExit();
247 | inFile.deleteOnExit();
248 |
249 | try {
250 | RandomAccessFile inRaf = new RandomAccessFile(inFile, "rw");
251 | byte[] opus = IOUtils.toByteArray(is);
252 | inRaf.write(opus);
253 |
254 | sampleRate = OggOpus.decode(inFilePath, outFilePath, sampleRate); // zero means to detect the sample rate by decoder
255 |
256 | RandomAccessFile outRaf = new RandomAccessFile(outFile, "r");
257 |
258 | byte[] data = new byte[(int)outRaf.length()];
259 |
260 | int outLength = outRaf.read(data);
261 |
262 | inRaf.close();
263 | outRaf.close();
264 | if(outLength == 0){
265 | throw new IOException("Data reading failed");
266 | }
267 | return data;
268 | } catch (FileNotFoundException e) {
269 | e.printStackTrace();
270 | } catch (IOException e) {
271 | e.printStackTrace();
272 | }
273 | return new byte[0];
274 | }
275 |
276 | /**
277 | * Analyze sample rate and return the PCM data
278 | * @param i
279 | * @return
280 | */
281 | public byte[] analyzeWavData(InputStream i){
282 | try {
283 | int headSize=44, metaDataSize=48;
284 | byte[] data = IOUtils.toByteArray(i);
285 | if(data.length < headSize){
286 | throw new IOException("Wrong Wav header");
287 | }
288 |
289 | if(this.sampleRate == 0 && data.length > 28) {
290 | this.sampleRate = readInt(data, 24); // 24 is the position of sample rate in wav format
291 | }
292 |
293 | int destPos = headSize + metaDataSize;
294 | int rawLength = data.length - destPos;
295 |
296 | byte[] d = new byte[rawLength];
297 | System.arraycopy(data, destPos, d, 0, rawLength);
298 | return d;
299 | } catch (IOException e) {
300 | Log.e(TAG, "Error while formatting header");
301 | }
302 | return new byte[0];
303 | }
304 |
305 | public byte[] stripHeaderAndSaveWav(InputStream i) {
306 | byte[] d = new byte[0];
307 | try {
308 | int headSize=44;
309 | int metaDataSize=48;
310 | i.skip(headSize+metaDataSize);
311 | d = IOUtils.toByteArray(i);
312 | } catch (IOException e) {
313 | Log.d(TAG,"Error while formatting header");
314 | }
315 | return saveWav(d);
316 | }
317 |
318 | public byte[] saveWav(byte[] d){
319 | PcmWaveWriter wR = new PcmWaveWriter(sampleRate, 1);
320 | return wR.saveWav(d, sampleRate, 1, 16);
321 | }
322 |
323 | void saveWavFile(byte[] d) {
324 | String fileName = getBaseDir() + "a.wav";
325 | try {
326 | PcmWaveWriter wR = new PcmWaveWriter(sampleRate, 1);
327 | wR.open(fileName);
328 | wR.saveWavFile(d,sampleRate,1,16);
329 | wR.close();
330 | Log.i(TAG, "save file OK");
331 | } catch (IOException e) {
332 | Log.d(TAG, "save file FAIL");
333 | e.printStackTrace();
334 | }
335 | }
336 |
337 | /**
338 | * Converts Little Endian (Windows) bytes to an int (Java uses Big Endian).
339 | * @param data the data to read.
340 | * @param offset the offset from which to start reading.
341 | * @return the integer value of the reassembled bytes.
342 | */
343 | protected static int readInt(final byte[] data, final int offset)
344 | {
345 | return (data[offset] & 0xff) |
346 | ((data[offset+1] & 0xff) << 8) |
347 | ((data[offset+2] & 0xff) << 16) |
348 | (data[offset+3] << 24); // no 0xff on the last one to keep the sign
349 | }
350 |
351 | /**
352 | * Converts Little Endian (Windows) bytes to an short (Java uses Big Endian).
353 | * @param data the data to read.
354 | * @param offset the offset from which to start reading.
355 | * @return the integer value of the reassembled bytes.
356 | */
357 | protected static int readShort(final byte[] data, final int offset)
358 | {
359 | return (data[offset] & 0xff) |
360 | (data[offset+1] << 8); // no 0xff on the last one to keep the sign
361 | }
362 | }
363 |
--------------------------------------------------------------------------------
/speech-android-wrapper/src/main/java/com/ibm/watson/developer_cloud/android/text_to_speech/v1/TextToSpeech.java:
--------------------------------------------------------------------------------
1 | /**
2 | * © Copyright IBM Corporation 2015
3 | *
4 | * Licensed under the Apache License, Version 2.0 (the "License");
5 | * you may not use this file except in compliance with the License.
6 | * You may obtain a copy of the License at
7 | *
8 | * http://www.apache.org/licenses/LICENSE-2.0
9 | *
10 | * Unless required by applicable law or agreed to in writing, software
11 | * distributed under the License is distributed on an "AS IS" BASIS,
12 | * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13 | * See the License for the specific language governing permissions and
14 | * limitations under the License.
15 | **/
16 |
17 | package com.ibm.watson.developer_cloud.android.text_to_speech.v1;
18 |
19 | import android.util.Log;
20 |
21 | import com.ibm.watson.developer_cloud.android.speech_common.v1.TokenProvider;
22 |
23 | import org.apache.http.HttpResponse;
24 | import org.apache.http.auth.UsernamePasswordCredentials;
25 | import org.apache.http.client.HttpClient;
26 | import org.apache.http.client.methods.HttpGet;
27 | import org.apache.http.impl.auth.BasicScheme;
28 | import org.apache.http.impl.client.DefaultHttpClient;
29 |
30 | import org.json.JSONException;
31 | import org.json.JSONObject;
32 |
33 | import java.io.BufferedReader;
34 | import java.io.IOException;
35 | import java.io.InputStream;
36 | import java.io.InputStreamReader;
37 | import java.net.URI;
38 |
39 | /**
40 | * Speech Recognition Class for SDK functions
41 | * @author Viney Ugave (vaugave@us.ibm.com)
42 | */
43 | public class TextToSpeech {
44 | protected static final String TAG = "TextToSpeech";
45 |
46 | private TTSUtility ttsUtility;
47 | private String username;
48 | private String password;
49 | private URI hostURL;
50 | private TokenProvider tokenProvider = null;
51 | private String voice;
52 | private boolean learningOptOut = false;
53 |
54 | /**Speech Recognition Shared Instance
55 | *
56 | */
57 | private static TextToSpeech _instance = null;
58 |
59 | public static TextToSpeech sharedInstance(){
60 | if(_instance == null){
61 | synchronized(TextToSpeech.class){
62 | _instance = new TextToSpeech();
63 | }
64 | }
65 | return _instance;
66 | }
67 |
68 | /**
69 | * Init the shared instance with the context
70 | * @param uri
71 | */
72 | public void initWithContext(URI uri){
73 | this.setHostURL(uri);
74 | }
75 |
76 | /**
77 | * Send request of TTS
78 | * @param ttsString
79 | */
80 | public void synthesize(String ttsString) {
81 | Log.d(TAG, "synthesize called: " + this.hostURL.toString() + "/v1/synthesize");
82 | String[] Arguments = {
83 | this.hostURL.toString()+"/v1/synthesize",
84 | this.username,
85 | this.password,
86 | this.voice,
87 | ttsString,
88 | this.tokenProvider == null ? null : this.tokenProvider.getToken(),
89 | this.learningOptOut ? "true" : null
90 | };
91 | try {
92 | ttsUtility = new TTSUtility();
93 | ttsUtility.setCodec(TTSUtility.CODEC_WAV);
94 | ttsUtility.synthesize(Arguments);
95 | }
96 | catch (Exception e) {
97 | e.printStackTrace();
98 | }
99 | }
100 |
101 | private void buildAuthenticationHeader(HttpGet httpGet) {
102 |
103 | // use token based authentication if possible, otherwise Basic Authentication will be used
104 | if (this.tokenProvider != null) {
105 | Log.d(TAG, "using token based authentication");
106 | httpGet.setHeader("X-Watson-Authorization-Token",this.tokenProvider.getToken());
107 | } else {
108 | Log.d(TAG, "using basic authentication");
109 | httpGet.setHeader(BasicScheme.authenticate(new UsernamePasswordCredentials(this.username, this.password), "UTF-8",false));
110 | }
111 | }
112 |
113 | public JSONObject getVoices() {
114 |
115 | JSONObject object = null;
116 |
117 | try {
118 | HttpClient httpClient = new DefaultHttpClient();
119 | HttpGet httpGet = new HttpGet(this.hostURL+"/v1/voices");
120 | Log.d(TAG,"url: " + this.hostURL+"/v1/voices");
121 | this.buildAuthenticationHeader(httpGet);
122 | httpGet.setHeader("accept", "application/json");
123 | HttpResponse executed = httpClient.execute(httpGet);
124 | InputStream is = executed.getEntity().getContent();
125 |
126 | // get the JSON object containing the models from the InputStream
127 | BufferedReader streamReader = new BufferedReader(new InputStreamReader(is, "UTF-8"));
128 | StringBuilder responseStrBuilder = new StringBuilder();
129 | String inputStr;
130 | while ((inputStr = streamReader.readLine()) != null)
131 | responseStrBuilder.append(inputStr);
132 | object = new JSONObject(responseStrBuilder.toString());
133 | Log.d(TAG, object.toString());
134 | } catch (IOException | JSONException e) {
135 | e.printStackTrace();
136 | }
137 |
138 | return object;
139 | }
140 |
141 | /**
142 | * Set credentials
143 | * @param username
144 | * @param password
145 | */
146 | public void setCredentials(String username, String password) {
147 | this.username = username;
148 | this.password = password;
149 | }
150 |
151 | /**
152 | * Set host URL
153 | * @param hostURL
154 | */
155 | public void setHostURL(URI hostURL) {
156 | this.hostURL = hostURL;
157 | }
158 |
159 | /**
160 | * Set token provider (for token based authentication)
161 | */
162 | public void setTokenProvider(TokenProvider tokenProvider) { this.tokenProvider = tokenProvider; }
163 |
164 | /**
165 | * Set TTS voice
166 | */
167 | public void setVoice(String voice) {
168 | this.voice = voice;
169 | }
170 |
171 | /**
172 | * Set X-Watson-Learning-OptOut
173 | * @param optOut
174 | */
175 | public void setLearningOptOut(boolean optOut) { this.learningOptOut = optOut; }
176 | }
177 |
--------------------------------------------------------------------------------
/speech-android-wrapper/src/main/jniLibs/arm64-v8a/libjnidispatch.so:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/watson-developer-cloud/speech-android-sdk/149f105c3e82c40b07e49f449fbf54b602de614d/speech-android-wrapper/src/main/jniLibs/arm64-v8a/libjnidispatch.so
--------------------------------------------------------------------------------
/speech-android-wrapper/src/main/jniLibs/arm64-v8a/libogg.so:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/watson-developer-cloud/speech-android-sdk/149f105c3e82c40b07e49f449fbf54b602de614d/speech-android-wrapper/src/main/jniLibs/arm64-v8a/libogg.so
--------------------------------------------------------------------------------
/speech-android-wrapper/src/main/jniLibs/arm64-v8a/liboggopus.so:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/watson-developer-cloud/speech-android-sdk/149f105c3e82c40b07e49f449fbf54b602de614d/speech-android-wrapper/src/main/jniLibs/arm64-v8a/liboggopus.so
--------------------------------------------------------------------------------
/speech-android-wrapper/src/main/jniLibs/arm64-v8a/libopus.so:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/watson-developer-cloud/speech-android-sdk/149f105c3e82c40b07e49f449fbf54b602de614d/speech-android-wrapper/src/main/jniLibs/arm64-v8a/libopus.so
--------------------------------------------------------------------------------
/speech-android-wrapper/src/main/jniLibs/arm64-v8a/libspeex.so:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/watson-developer-cloud/speech-android-sdk/149f105c3e82c40b07e49f449fbf54b602de614d/speech-android-wrapper/src/main/jniLibs/arm64-v8a/libspeex.so
--------------------------------------------------------------------------------
/speech-android-wrapper/src/main/jniLibs/arm64-v8a/libspeexdsp.so:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/watson-developer-cloud/speech-android-sdk/149f105c3e82c40b07e49f449fbf54b602de614d/speech-android-wrapper/src/main/jniLibs/arm64-v8a/libspeexdsp.so
--------------------------------------------------------------------------------
/speech-android-wrapper/src/main/jniLibs/armeabi-v7a/libjnidispatch.so:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/watson-developer-cloud/speech-android-sdk/149f105c3e82c40b07e49f449fbf54b602de614d/speech-android-wrapper/src/main/jniLibs/armeabi-v7a/libjnidispatch.so
--------------------------------------------------------------------------------
/speech-android-wrapper/src/main/jniLibs/armeabi-v7a/libogg.so:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/watson-developer-cloud/speech-android-sdk/149f105c3e82c40b07e49f449fbf54b602de614d/speech-android-wrapper/src/main/jniLibs/armeabi-v7a/libogg.so
--------------------------------------------------------------------------------
/speech-android-wrapper/src/main/jniLibs/armeabi-v7a/liboggopus.so:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/watson-developer-cloud/speech-android-sdk/149f105c3e82c40b07e49f449fbf54b602de614d/speech-android-wrapper/src/main/jniLibs/armeabi-v7a/liboggopus.so
--------------------------------------------------------------------------------
/speech-android-wrapper/src/main/jniLibs/armeabi-v7a/libopus.so:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/watson-developer-cloud/speech-android-sdk/149f105c3e82c40b07e49f449fbf54b602de614d/speech-android-wrapper/src/main/jniLibs/armeabi-v7a/libopus.so
--------------------------------------------------------------------------------
/speech-android-wrapper/src/main/jniLibs/armeabi-v7a/libspeex.so:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/watson-developer-cloud/speech-android-sdk/149f105c3e82c40b07e49f449fbf54b602de614d/speech-android-wrapper/src/main/jniLibs/armeabi-v7a/libspeex.so
--------------------------------------------------------------------------------
/speech-android-wrapper/src/main/jniLibs/armeabi-v7a/libspeexdsp.so:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/watson-developer-cloud/speech-android-sdk/149f105c3e82c40b07e49f449fbf54b602de614d/speech-android-wrapper/src/main/jniLibs/armeabi-v7a/libspeexdsp.so
--------------------------------------------------------------------------------
/speech-android-wrapper/src/main/jniLibs/armeabi/libjnidispatch.so:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/watson-developer-cloud/speech-android-sdk/149f105c3e82c40b07e49f449fbf54b602de614d/speech-android-wrapper/src/main/jniLibs/armeabi/libjnidispatch.so
--------------------------------------------------------------------------------
/speech-android-wrapper/src/main/jniLibs/armeabi/libogg.so:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/watson-developer-cloud/speech-android-sdk/149f105c3e82c40b07e49f449fbf54b602de614d/speech-android-wrapper/src/main/jniLibs/armeabi/libogg.so
--------------------------------------------------------------------------------
/speech-android-wrapper/src/main/jniLibs/armeabi/liboggopus.so:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/watson-developer-cloud/speech-android-sdk/149f105c3e82c40b07e49f449fbf54b602de614d/speech-android-wrapper/src/main/jniLibs/armeabi/liboggopus.so
--------------------------------------------------------------------------------
/speech-android-wrapper/src/main/jniLibs/armeabi/libopus.so:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/watson-developer-cloud/speech-android-sdk/149f105c3e82c40b07e49f449fbf54b602de614d/speech-android-wrapper/src/main/jniLibs/armeabi/libopus.so
--------------------------------------------------------------------------------
/speech-android-wrapper/src/main/jniLibs/armeabi/libspeex.so:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/watson-developer-cloud/speech-android-sdk/149f105c3e82c40b07e49f449fbf54b602de614d/speech-android-wrapper/src/main/jniLibs/armeabi/libspeex.so
--------------------------------------------------------------------------------
/speech-android-wrapper/src/main/jniLibs/armeabi/libspeexdsp.so:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/watson-developer-cloud/speech-android-sdk/149f105c3e82c40b07e49f449fbf54b602de614d/speech-android-wrapper/src/main/jniLibs/armeabi/libspeexdsp.so
--------------------------------------------------------------------------------
/speech-android-wrapper/src/main/jniLibs/mips/libjnidispatch.so:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/watson-developer-cloud/speech-android-sdk/149f105c3e82c40b07e49f449fbf54b602de614d/speech-android-wrapper/src/main/jniLibs/mips/libjnidispatch.so
--------------------------------------------------------------------------------
/speech-android-wrapper/src/main/jniLibs/mips/libogg.so:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/watson-developer-cloud/speech-android-sdk/149f105c3e82c40b07e49f449fbf54b602de614d/speech-android-wrapper/src/main/jniLibs/mips/libogg.so
--------------------------------------------------------------------------------
/speech-android-wrapper/src/main/jniLibs/mips/liboggopus.so:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/watson-developer-cloud/speech-android-sdk/149f105c3e82c40b07e49f449fbf54b602de614d/speech-android-wrapper/src/main/jniLibs/mips/liboggopus.so
--------------------------------------------------------------------------------
/speech-android-wrapper/src/main/jniLibs/mips/libopus.so:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/watson-developer-cloud/speech-android-sdk/149f105c3e82c40b07e49f449fbf54b602de614d/speech-android-wrapper/src/main/jniLibs/mips/libopus.so
--------------------------------------------------------------------------------
/speech-android-wrapper/src/main/jniLibs/mips/libspeex.so:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/watson-developer-cloud/speech-android-sdk/149f105c3e82c40b07e49f449fbf54b602de614d/speech-android-wrapper/src/main/jniLibs/mips/libspeex.so
--------------------------------------------------------------------------------
/speech-android-wrapper/src/main/jniLibs/mips/libspeexdsp.so:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/watson-developer-cloud/speech-android-sdk/149f105c3e82c40b07e49f449fbf54b602de614d/speech-android-wrapper/src/main/jniLibs/mips/libspeexdsp.so
--------------------------------------------------------------------------------
/speech-android-wrapper/src/main/jniLibs/mips64/libjnidispatch.so:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/watson-developer-cloud/speech-android-sdk/149f105c3e82c40b07e49f449fbf54b602de614d/speech-android-wrapper/src/main/jniLibs/mips64/libjnidispatch.so
--------------------------------------------------------------------------------
/speech-android-wrapper/src/main/jniLibs/mips64/libogg.so:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/watson-developer-cloud/speech-android-sdk/149f105c3e82c40b07e49f449fbf54b602de614d/speech-android-wrapper/src/main/jniLibs/mips64/libogg.so
--------------------------------------------------------------------------------
/speech-android-wrapper/src/main/jniLibs/mips64/liboggopus.so:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/watson-developer-cloud/speech-android-sdk/149f105c3e82c40b07e49f449fbf54b602de614d/speech-android-wrapper/src/main/jniLibs/mips64/liboggopus.so
--------------------------------------------------------------------------------
/speech-android-wrapper/src/main/jniLibs/mips64/libopus.so:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/watson-developer-cloud/speech-android-sdk/149f105c3e82c40b07e49f449fbf54b602de614d/speech-android-wrapper/src/main/jniLibs/mips64/libopus.so
--------------------------------------------------------------------------------
/speech-android-wrapper/src/main/jniLibs/mips64/libspeex.so:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/watson-developer-cloud/speech-android-sdk/149f105c3e82c40b07e49f449fbf54b602de614d/speech-android-wrapper/src/main/jniLibs/mips64/libspeex.so
--------------------------------------------------------------------------------
/speech-android-wrapper/src/main/jniLibs/mips64/libspeexdsp.so:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/watson-developer-cloud/speech-android-sdk/149f105c3e82c40b07e49f449fbf54b602de614d/speech-android-wrapper/src/main/jniLibs/mips64/libspeexdsp.so
--------------------------------------------------------------------------------
/speech-android-wrapper/src/main/jniLibs/x86/libjnidispatch.so:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/watson-developer-cloud/speech-android-sdk/149f105c3e82c40b07e49f449fbf54b602de614d/speech-android-wrapper/src/main/jniLibs/x86/libjnidispatch.so
--------------------------------------------------------------------------------
/speech-android-wrapper/src/main/jniLibs/x86/libogg.so:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/watson-developer-cloud/speech-android-sdk/149f105c3e82c40b07e49f449fbf54b602de614d/speech-android-wrapper/src/main/jniLibs/x86/libogg.so
--------------------------------------------------------------------------------
/speech-android-wrapper/src/main/jniLibs/x86/liboggopus.so:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/watson-developer-cloud/speech-android-sdk/149f105c3e82c40b07e49f449fbf54b602de614d/speech-android-wrapper/src/main/jniLibs/x86/liboggopus.so
--------------------------------------------------------------------------------
/speech-android-wrapper/src/main/jniLibs/x86/libopus.so:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/watson-developer-cloud/speech-android-sdk/149f105c3e82c40b07e49f449fbf54b602de614d/speech-android-wrapper/src/main/jniLibs/x86/libopus.so
--------------------------------------------------------------------------------
/speech-android-wrapper/src/main/jniLibs/x86/libspeex.so:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/watson-developer-cloud/speech-android-sdk/149f105c3e82c40b07e49f449fbf54b602de614d/speech-android-wrapper/src/main/jniLibs/x86/libspeex.so
--------------------------------------------------------------------------------
/speech-android-wrapper/src/main/jniLibs/x86/libspeexdsp.so:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/watson-developer-cloud/speech-android-sdk/149f105c3e82c40b07e49f449fbf54b602de614d/speech-android-wrapper/src/main/jniLibs/x86/libspeexdsp.so
--------------------------------------------------------------------------------
/speech-android-wrapper/src/main/jniLibs/x86_64/libjnidispatch.so:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/watson-developer-cloud/speech-android-sdk/149f105c3e82c40b07e49f449fbf54b602de614d/speech-android-wrapper/src/main/jniLibs/x86_64/libjnidispatch.so
--------------------------------------------------------------------------------
/speech-android-wrapper/src/main/jniLibs/x86_64/libogg.so:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/watson-developer-cloud/speech-android-sdk/149f105c3e82c40b07e49f449fbf54b602de614d/speech-android-wrapper/src/main/jniLibs/x86_64/libogg.so
--------------------------------------------------------------------------------
/speech-android-wrapper/src/main/jniLibs/x86_64/liboggopus.so:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/watson-developer-cloud/speech-android-sdk/149f105c3e82c40b07e49f449fbf54b602de614d/speech-android-wrapper/src/main/jniLibs/x86_64/liboggopus.so
--------------------------------------------------------------------------------
/speech-android-wrapper/src/main/jniLibs/x86_64/libopus.so:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/watson-developer-cloud/speech-android-sdk/149f105c3e82c40b07e49f449fbf54b602de614d/speech-android-wrapper/src/main/jniLibs/x86_64/libopus.so
--------------------------------------------------------------------------------
/speech-android-wrapper/src/main/jniLibs/x86_64/libspeex.so:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/watson-developer-cloud/speech-android-sdk/149f105c3e82c40b07e49f449fbf54b602de614d/speech-android-wrapper/src/main/jniLibs/x86_64/libspeex.so
--------------------------------------------------------------------------------
/speech-android-wrapper/src/main/jniLibs/x86_64/libspeexdsp.so:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/watson-developer-cloud/speech-android-sdk/149f105c3e82c40b07e49f449fbf54b602de614d/speech-android-wrapper/src/main/jniLibs/x86_64/libspeexdsp.so
--------------------------------------------------------------------------------
/speech-android-wrapper/src/main/res/drawable-hdpi/ic_launcher.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/watson-developer-cloud/speech-android-sdk/149f105c3e82c40b07e49f449fbf54b602de614d/speech-android-wrapper/src/main/res/drawable-hdpi/ic_launcher.png
--------------------------------------------------------------------------------
/speech-android-wrapper/src/main/res/drawable-mdpi/ic_launcher.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/watson-developer-cloud/speech-android-sdk/149f105c3e82c40b07e49f449fbf54b602de614d/speech-android-wrapper/src/main/res/drawable-mdpi/ic_launcher.png
--------------------------------------------------------------------------------
/speech-android-wrapper/src/main/res/drawable-xhdpi/ic_launcher.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/watson-developer-cloud/speech-android-sdk/149f105c3e82c40b07e49f449fbf54b602de614d/speech-android-wrapper/src/main/res/drawable-xhdpi/ic_launcher.png
--------------------------------------------------------------------------------
/speech-android-wrapper/src/main/res/mipmap-hdpi/ic_launcher.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/watson-developer-cloud/speech-android-sdk/149f105c3e82c40b07e49f449fbf54b602de614d/speech-android-wrapper/src/main/res/mipmap-hdpi/ic_launcher.png
--------------------------------------------------------------------------------
/speech-android-wrapper/src/main/res/mipmap-mdpi/ic_launcher.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/watson-developer-cloud/speech-android-sdk/149f105c3e82c40b07e49f449fbf54b602de614d/speech-android-wrapper/src/main/res/mipmap-mdpi/ic_launcher.png
--------------------------------------------------------------------------------
/speech-android-wrapper/src/main/res/mipmap-xhdpi/ic_launcher.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/watson-developer-cloud/speech-android-sdk/149f105c3e82c40b07e49f449fbf54b602de614d/speech-android-wrapper/src/main/res/mipmap-xhdpi/ic_launcher.png
--------------------------------------------------------------------------------
/speech-android-wrapper/src/main/res/mipmap-xxhdpi/ic_launcher.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/watson-developer-cloud/speech-android-sdk/149f105c3e82c40b07e49f449fbf54b602de614d/speech-android-wrapper/src/main/res/mipmap-xxhdpi/ic_launcher.png
--------------------------------------------------------------------------------
/speech-android-wrapper/src/main/res/mipmap-xxxhdpi/ic_launcher.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/watson-developer-cloud/speech-android-sdk/149f105c3e82c40b07e49f449fbf54b602de614d/speech-android-wrapper/src/main/res/mipmap-xxxhdpi/ic_launcher.png
--------------------------------------------------------------------------------
/speech-android-wrapper/src/main/res/values-v11/styles.xml:
--------------------------------------------------------------------------------
1 |
2 |
3 |
7 |
10 |
11 |
12 |
--------------------------------------------------------------------------------
/speech-android-wrapper/src/main/res/values-v14/styles.xml:
--------------------------------------------------------------------------------
1 |
2 |
3 |
8 |
11 |
12 |
13 |
--------------------------------------------------------------------------------
/speech-android-wrapper/src/main/res/values/styles.xml:
--------------------------------------------------------------------------------
1 |
2 |
3 |
7 |
14 |
15 |
16 |
19 |
20 |
21 |
--------------------------------------------------------------------------------