Appendix I - MATLAB Programs

The procedures outlined in these notes, should be combined into MATLAB programs.For more on M-file programming we refer the reader to the MATLAB user guide. The following are two M-files which combined do the tests outlined above.

% M-file program that reads the two voices of 'test1' and 'test2' and computes
% their average fourier transform
function z=meanvoice
y1=wavread('test1');
y2=wavread('test2');
x1=fft(y1,60000);
x2=fft(y2,60000);
z1=x1(1:2500);
z2=x2(1:2500);
z1(1:500)=zeros(500,1);
z2(1:500)=zeros(500,1);
n1=sqrt(sum(abs(z1).^2));
n2=sqrt(sum(abs(z2).^2));
z1=z1/n1;
z2=z2/n2;
z=(z1+z2)/2;
n=sqrt(sum(abs(z).^2));
z=z/n;

By typing

Z=meanvoice

One obtains the average Fourier Trans form. The next tests a third voice against this average;

%A MATLAB fuction which tests the voice against
%a reference.
function test(Z)
name=input('filename? ','s');
y=wavread(name);
%x is the Fourier transform of y
x=fft(y,60000);
z=x(1:2500);
z(1:500)=zeros(500,1);
s=sqrt(sum(abs(z-Z).^2));
if s< 1
    'OK'
else
    'IMPOSTOR!'
end

Typing

test(Z)

the program will use the average computed before and prompt the user to give the name of a WAV-file which contains the voice to be tested. It then performs the necessary operations and tests this voice against the average Z.