lin_sep.m

实现多输入的线性混合信号的分离,具体使用方法见readme.txt,readme.txt中的部分程序因某些原因

function [ys,W,Q]=Lin_Sep(x,Fs);
%  The main program of maximum SNR algorithm

%  Usage:  [ys,W,Q]=Lin_Sep(x,Fs);

%  This m file is just used in the early stage of the test, however, it is
%  no longer necessary if we add the tranposing part of vectors to the
%  SNR_Max.m 

%  Input:
%  *x is a column vectors of signals received by the senor or mixed signals 
%  *Fs is the sampling rate

%  Output:
%  *ys is a column vectors of reconstructed signals or demixed signals
%  *W is the demixing matrix
%  *Q is the whitening matrix



%  x=wavread('X_linear.wav');% To further improve this program  I suppose to
%  involve the 'Voice Input' part in my program
x=x'; 
[ys,W,Q]=SNR_Max(x,Fs);
ys=ys';

%---------------Experiment Results With Gaussian White Noise------------%
% SNR=60dB Sense no Crosstalk
% SNR=50dB Sense no Crosstalk 
% SNR=20dB Slight Crosstalk
% SNR=10dB Y110 is recognizable while Y210 has recognizable crosstalk
% SNR=00dB Y100 is also recognizable while Y200 has been heavy disturbed
% SNR=-05dB Y1m05 is also recognizable while Y2m05 has been heavy disturbed
% SNR=-10dB Y1m10 has totally been submerged in noice; Amplified Y2m10 is
% recognizable but with loud noise and slight crosstalk

%----------------------------Wow!! I solved it!!------------------------------%
%This is the first time I felt so exciting and want to tell everyone that I
%have solved a problem puzzled me for a long time, all the efforts were reduced to
%so beautiful a general matrix!! And a sentence given by Helen Keller suddenly 
%come up to my mind:I knew there were barriers still but barries could in time 
%be swept away! OK after a short period of excitement, I am setting out again 
%toward another peak in my life