How Can I Inverse A Signal?

Question

1.00/5 (3 votes)

See more:

Hi all
I want to run my code that is about Speech Enhancement Based on Deep Neural Networks and our steps are

1-framing signal * hamming window
2-log-power spectral frame
3-Deep learning
4-inverse output to signal

So my problem is inversing can not return true answer my code

VB

function [ out ] = dfft( s )


 N=300;
 M=20;
 L=length(s);

 NumFram=floor(L/(N-M))-1;
 frames=zeros(NumFram,N);
 for i=1:N

 frames(1,i)=s(i);

 end
 j=1;


 for j=2:NumFram
 n=1;

 for i= (j-1) * (N-M) : (j-1) * (N-M) + N

 frames(j,n)=s(i);
 n=n+1;


 end
 end

 frames(:,N+1)=[];


 HamiW=zeros(1,N);
 HamW=hamming(N);

for i=1:N
HamiW(1,i)=HamW(i);
end


Y=zeros(NumFram,N);

Y=frames*diag(HamiW);

 FFT=zeros(NumFram,N);


 for i=1:NumFram

 FFT(i,:)=fft(Y(i,:));
 for j=1:N
 phase(i,j)=atan(imag(FFT(i,j))/real(FFT(i,j)));
 logfft(i,j)=FFT(i,j)*conj(FFT(i,j));
 Z=sqrt(logfft(i,j));
 S(i,j)= complex(Z*cos(phase(i,j)),Z*sin(phase(i,j)));


 end
out(i,:)=ifft(S(i,:));

end

Posted 26-Jul-15 0:16am

Sa.Mo

Updated 30-Jul-15 3:27am

v3

Add a Solution

Comments

Kenneth Haugland 26-Jul-15 7:52am

Usually the power spectrum is 20*log10(abs(FFT)), you seem to favor the natural logarithm? In any cast that is just for viewing the spectrum.

I don't know what you mean by inverse the signal, if it is back to its original you need to perform an ifft of the fft.

Sa.Mo 30-Jul-15 9:28am

can u check my update code? thanks

Sergey Alexandrovich Kryukov 27-Jul-15 0:36am

Multiply the signal by −1...
—SA

Sa.Mo 31-Jul-15 8:02am

was not work..

1 solution

Add a Solution

Add your solution here

Treat my content as plain text, not as HTML

Preview 0

…

Existing Members

Sign in to your account

...or Join us

Download, Vote, Comment, Publish.

Your Email
Password
Forgot your password?

Your Email
This email is in use. Do you need your password?
Optional Password

I have read and agree to the Terms of Service and Privacy Policy
Please subscribe me to the CodeProject newsletters

When answering a question please:

Read the question carefully.
Understand that English isn't everyone's first language so be lenient of bad spelling and grammar.
If a question is poorly phrased then either ask for clarification, ignore it, or edit the question and fix the problem. Insults are not welcome.
Don't tell someone to read the manual. Chances are they have and don't get it. Provide an answer or move on to the next question.

Let's work to help developers, not make them feel stupid.

This content, along with any associated source code and files, is licensed under The Code Project Open License (CPOL)

Kenneth Haugland · Answer 1 · 2015-07-30T06:10:00

You seem to lack some knowledge on DSP (Digital Signal Prepossessing) so I would suggest you start reading some books or online material to gain some background information on what you are working with here.

Normally we take the Hamming[^], Blackman or some other window function to filter input signal with voice in it. It is basically a specialty designed filter. And it is normally used together with convolution of signals[^].
If memory serves me correctly it was used the following way (assumeing that f(t) is the time dependent signal and Ham(t) is the Hamming window):

C#

output = conv(f,Ham);

Once you have your output signal you can take the fft of it to get the spectrum. It is normal to take the log of the signal to get more human readable signal.To plot it together with a frequency you need to know the sample rate as well. Taken form this[^]:

C++

nf=1024; %number of point in DTFT   
Y = fft(y,nf);    
f = fs/2*linspace(0,1,nf/2+1);    
plot(f,log(1 + abs(Y(1:nf/2+1)))); %// Change    
title('Single-Sided Amplitude Spectrum of y(t)')   
xlabel('Frequency (Hz)')   
ylabel('|Y(f)|')

Now to understand what you do when you actually take an fft of a signal, that you transform it to the frequency domain from the time domain. You can also go the other way from the frequency domain to the time domain with an ifft. However some information is lost on the way, so what you get is the Impulse Response, and not the original signal. So information got lost on the way.