% G. Dempsey
% Bradley University
% 2/23/2009
% delete W,b initialization to zero, newlin does this
% Neural Networks: HW12.m

clear

% get time domain info for 6 students

y1=wavread('olee.wav');		% open for Lee
y2=wavread('glee.wav');		% give me for Lee
y3=wavread('ojoel.wav');
y4=wavread('gjoel.wav');
y5=wavread('ojeff.wav');
y6=wavread('gjeff.wav');
y7=wavread('ocurtis.wav');
y8=wavread('gcurtis.wav');
y9=wavread('okurt.wav');
y10=wavread('gkurt.wav');
y11=wavread('obecky.wav');
y12=wavread('gbecky.wav');

% eliminate DC content

m1=mean(y1);m2=mean(y2);m3=mean(y3);
m4=mean(y4);m5=mean(y5);m6=mean(y6);
m7=mean(y7);m8=mean(y8);m9=mean(y9);
m10=mean(y10);m11=mean(y11);m12=mean(y12);

y1=y1-m1;y2=y2-m2;y3=y3-m3;y4=y4-m4;
y5=y5-m5;y6=y6-m6;y7=y7-m7;y8=y8-m8;
y9=y9-m9;y10=y10-m10;y11=y11-m11;
y12=y12-m12;

% normalize data

y1=y1/max(abs(y1));y2=y2/max(abs(y2));
y3=y3/max(abs(y3));y4=y4/max(abs(y4));
y5=y5/max(abs(y5));y6=y6/max(abs(y6));
y7=y7/max(abs(y7));y8=y8/max(abs(y8));
y9=y9/max(abs(y9));y10=y10/max(abs(y10));
y11=y11/max(abs(y11));y12=y12/max(abs(y12));

disp('Compute FFT- press any key')
pause

% setup code for frequency response of input signals (fft)
Npts=256;		% 256 pt FFT to distinquish freuencies
			% freq div= 8KHz/256= 31.25 Hz
x1=fft(y1,Npts);	% take FFT
x1=abs(x1);		% take absolute value
			% normalize data 
x1=x1/max(x1);

% do other voices

x2=fft(y2,Npts);x2=abs(x2);x2=x2/max(x2);

x3=fft(y3,Npts);x3=abs(x3);x3=x3/max(x3);

x4=fft(y4,Npts);x4=abs(x4);x4=x4/max(x4);

x5=fft(y5,Npts);x5=abs(x5);x5=x5/max(x5);

x6=fft(y6,Npts);x6=abs(x6);x6=x6/max(x6);

x7=fft(y7,Npts);x7=abs(x7);x7=x7/max(x7);

x8=fft(y8,Npts);x8=abs(x8);x8=x8/max(x8);

x9=fft(y9,Npts);x9=abs(x9);x9=x9/max(x9);

x10=fft(y10,Npts);x10=abs(x10);x10=x10/max(x10);

x11=fft(y11,Npts);x11=abs(x11);x11=x11/max(x11);

x12=fft(y12,Npts);x12=abs(x12);x12=x12/max(x12);

% save data for ANN training

x1=x1(1:48);		% 0 to 1500 Hz only (48 inputs)
x2=x2(1:48);x3=x3(1:48);x4=x4(1:48);x5=x5(1:48);
x6=x6(1:48);x7=x7(1:48);x8=x8(1:48);x9=x9(1:48);
x10=x10(1:48);x11=x11(1:48);x12=x12(1:48);

disp('Press any key to start training')
pause

% Define input pattern

P = [x1 x2 x3 x4 x5 x6 x7 x8 x9 x10 x11 x12];

% Define Targets:

T = [1 -1 1 -1 1 -1 1 -1 1 -1 1 -1];

% build a matrix (2,48) with column 1 all 0's (min value of input components)
% and column 2 all 1's (max value of input components)
c1=zeros(48,1);
c2=ones(48,1);
PR=[c1 c2];

% Setup the Network
% net = newlin(PR,S,ID,LR)
% 
% 	  NEWLIN(PR,S,ID,LR) takes these arguments,
% 	    PR - Rx2 matrix of min and max values for R input elements.
% 	    S  - Number of elements in the output vector.
% 	    ID - Input delay vector, default = [0].
% 	    LR - Learning rate, default = 0.01;
% 	  and returns a new linear layer.
%   default: weights, bias = 0

net=newlin(PR,1,0,0.005);

%    TRAINING THE NETWORK
%    ====================

%    Widrow-Hoff rule
net.trainParam.goal = 0.0;	% Sum-squared error goal.
net.trainParam.epochs = 1000;	% set maximum epochs
net.performFcn='sse';

%    Training begins...please wait...

[net,tr] = train(net,P,T);

%    ...and finishes.
disp('end of training')

% Display Training Results

disp('press any key to see actual outputs for 12 desired targets')
pause

A=sim(net,P)

% Display Test Set Results
disp('press any key to see test set outputs')
% get time domain info for next 2 test subjects

y13=wavread('ogary.wav');
y14=wavread('ggary.wav');
y15=wavread('odana.wav');
y16=wavread('gdana.wav');

% eliminate DC content

m13=mean(y13);
m14=mean(y14);
m15=mean(y15);
m16=mean(y16);

y13=y13-m13;
y14=y14-m14;
y15=y15-m15;
y16=y16-m16;

% normalize data

y13=y13/max(abs(y13));
y14=y14/max(abs(y14));
y15=y15/max(abs(y15));
y16=y16/max(abs(y16));

disp('Compute FFT for test subjects')
pause

 
x13=fft(y13,Npts);
x13=abs(x13);
x13=x13/max(x13);

x14=fft(y14,Npts);
x14=abs(x14);
x14=x14/max(x14);

x15=fft(y15,Npts);
x15=abs(x15);
x15=x15/max(x15);

x16=fft(y16,Npts);
x16=abs(x16);
x16=x16/max(x16);

% save data for ANN testing

x13=x13(1:48);
x14=x14(1:48);
x15=x15(1:48);
x16=x16(1:48);

% now test ANN
 
PT=[x13 x14 x15 x16];
A2=sim(net,PT)
disp('end of testing')