├── README.md
├── f_CSP.m
└── main_CSP.m


/README.md:
--------------------------------------------------------------------------------
1 | # Common-Spatial-Pattern-Algorithm
2 | Common spatial pattern (CSP) is a mathematical procedure used in signal processing for separating a multivariate signal into additive subcomponents which have maximum differences in variance between two windows. This algorithm is mainly used in motor imagery based BCI for processing EEG data. 
3 | 


--------------------------------------------------------------------------------
/f_CSP.m:
--------------------------------------------------------------------------------
 1 | function [result] = f_CSP(varargin)
 2 | % This code is for calulating the projection matrix for CSP 
 3 | % Haider Raza, Intelligent System Research Center, University of Ulster, Northern Ireland, UK.
 4 | %     Raza-H@email.ulster.ac.uk
 5 | %     Date: 03-Oct-2014
 6 | 
 7 | % Input:
 8 |               
 9 | %        left:  left hand data
10 | %       right: right hand data
11 | % 
12 | % Output:
13 | %       left: Left hand data
14 | %       right: right hand data    
15 | 
16 |     if (nargin ~= 2)
17 |         disp('Must have 2 classes for CSP!')
18 |     end
19 |     
20 |     Rsum=0;
21 |     %finding the covariance of each class and composite covariance
22 |     for i = 1:nargin 
23 |         %mean here?
24 |         R{i} = ((varargin{i}*varargin{i}')/trace(varargin{i}*varargin{i}'));%instantiate me before the loop!
25 |         %Ramoser equation (2)
26 |         Rsum=Rsum+R{i};
27 |     end
28 |    
29 |     %   Find Eigenvalues and Eigenvectors of RC
30 |     %   Sort eigenvalues in descending order
31 |     [EVecsum,EValsum] = eig(Rsum);
32 |     [EValsum,ind] = sort(diag(EValsum),'descend');
33 |     EVecsum = EVecsum(:,ind);
34 |     
35 |     %   Find Whitening Transformation Matrix - Ramoser Equation (3)
36 |         W = sqrt(inv(diag(EValsum))) * EVecsum';
37 |     
38 |     
39 |     for k = 1:nargin
40 |         S{k} = W * R{k} * W'; %       Whiten Data Using Whiting Transform - Ramoser Equation (4)
41 |     end
42 |    
43 |     %generalized eigenvectors/values
44 |     [B,D] = eig(S{1},S{2});
45 |     % Simultanous diagonalization
46 | 			% Should be equivalent to [B,D]=eig(S{1});
47 |     
48 |     [D,ind]=sort(diag(D));B=B(:,ind);
49 |     
50 |     %Resulting Projection Matrix-these are the spatial filter coefficients
51 |     result = B'*W;
52 | end
53 | 


--------------------------------------------------------------------------------
/main_CSP.m:
--------------------------------------------------------------------------------
 1 | % Call the CSP function using the left and right hand data
 2 |     % Input
 3 |     % C1_Data=Class 1 Data in Channels*Observations
 4 |     % C2_Data=Class 2 Data in Channels*Observations
 5 | 
 6 |     
 7 |    % Output: 
 8 |    % W=Channels*Channels  % Projection Matrix 
 9 |    
10 | [W] = f_CSP(C1_Data,C2_Data);
11 | 


--------------------------------------------------------------------------------