Prev: Hough Transform
Next: synchonous machine
From: maxroucool mvjz on 6 Nov 2009 05:40
Hi all,

I have to analyze a noise signal, so the first step is to suppress the noise of the from an audio signal. So I have two questions:
- First I am trying to use the wden() function, but there are so many different arguments, that I don’t know how to choose options.
Here is how to use the function: [XD,CXD,LXD] = wden(X,TPTR,SORH,SCAL,N,'wname'). But do you have more information about the arguments? Because, I don’t even know what do they represent. I try to compare the noise suppression with different parameters, but it is quite hard to make the difference. Do you know for example, which wname I should use with an audio signal?
- My second question is, is there other methods to suppress the noise? Like using a low-pass filter to suppress high frequencies that are in certain case not useful.

Thank you very much for your help.
+++
From: Wayne King on 6 Nov 2009 09:34
"maxroucool mvjz" <maxroucool(a)yahoo.fr> wrote in message <hd0uel$bi8$1(a)fred.mathworks.com>...
> Hi all,
>
> I have to analyze a noise signal, so the first step is to suppress the noise of the from an audio signal. So I have two questions:
> - First I am trying to use the wden() function, but there are so many different arguments, that I don&#8217;t know how to choose options.
> Here is how to use the function: [XD,CXD,LXD] = wden(X,TPTR,SORH,SCAL,N,'wname'). But do you have more information about the arguments? Because, I don&#8217;t even know what do they represent. I try to compare the noise suppression with different parameters, but it is quite hard to make the difference. Do you know for example, which wname I should use with an audio signal?
> - My second question is, is there other methods to suppress the noise? Like using a low-pass filter to suppress high frequencies that are in certain case not useful.
>
> Thank you very much for your help.
> +++

Hi, have you looked at the examples in the documentation for the wavelet denoising?

>>doc wden

There's really not a best answer to how to choose most of the input arguments in wden(). The answer is: it depends. I would suggest you go through the demo
denoisingsignalsdemo.m which addresses some of this with examples. You can access that demo by typing
>>edit denoisingsignalsdemo.m
at the command prompt or by Start -> Toolboxes -> Wavelet -> Demos

I think the answer to your second question is yes. In many cases simply filtering the signal may yield a very satisfactory denoised version. Again, it depends on a number of factors. If you can give people a little better idea of the kind of signals you are working on and your goals, someone could give you more specific help.

Wayne
From: maxroucool mvjz on 12 Nov 2009 12:12
"Wayne King" <wmkingty(a)gmail.com> wrote in message <hd1c4q$1pv$1(a)fred.mathworks.com>...
> "maxroucool mvjz" <maxroucool(a)yahoo.fr> wrote in message <hd0uel$bi8$1(a)fred.mathworks.com>...
> > Hi all,
> >
> > I have to analyze a noise signal, so the first step is to suppress the noise of the from an audio signal. So I have two questions:
> > - First I am trying to use the wden() function, but there are so many different arguments, that I don&#8217;t know how to choose options.
> > Here is how to use the function: [XD,CXD,LXD] = wden(X,TPTR,SORH,SCAL,N,'wname'). But do you have more information about the arguments? Because, I don&#8217;t even know what do they represent. I try to compare the noise suppression with different parameters, but it is quite hard to make the difference. Do you know for example, which wname I should use with an audio signal?
> > - My second question is, is there other methods to suppress the noise? Like using a low-pass filter to suppress high frequencies that are in certain case not useful.
> >
> > Thank you very much for your help.
> > +++
>
> Hi, have you looked at the examples in the documentation for the wavelet denoising?
>
> >>doc wden
>
> There's really not a best answer to how to choose most of the input arguments in wden(). The answer is: it depends. I would suggest you go through the demo
> denoisingsignalsdemo.m which addresses some of this with examples. You can access that demo by typing
> >>edit denoisingsignalsdemo.m
> at the command prompt or by Start -> Toolboxes -> Wavelet -> Demos
>
> I think the answer to your second question is yes. In many cases simply filtering the signal may yield a very satisfactory denoised version. Again, it depends on a number of factors. If you can give people a little better idea of the kind of signals you are working on and your goals, someone could give you more specific help.
>
> Wayne

Thank you Wayne for your answer.

I read it carefully and it was really useful! I am trying to compare different ways of filtering noise in audio signal.
About "simple filtering" of the signal, what should I filter? only really high frequencies (what is the limit?), or are there other things to do?

Thanks!
From: Wayne King on 13 Nov 2009 07:09
"maxroucool mvjz" <maxroucool(a)yahoo.fr> wrote in message <hdhfl2$akm$1(a)fred.mathworks.com>...
> "Wayne King" <wmkingty(a)gmail.com> wrote in message <hd1c4q$1pv$1(a)fred.mathworks.com>...
> > "maxroucool mvjz" <maxroucool(a)yahoo.fr> wrote in message <hd0uel$bi8$1(a)fred.mathworks.com>...
> > > Hi all,
> > >
> > > I have to analyze a noise signal, so the first step is to suppress the noise of the from an audio signal. So I have two questions:
> > > - First I am trying to use the wden() function, but there are so many different arguments, that I don&#8217;t know how to choose options.
> > > Here is how to use the function: [XD,CXD,LXD] = wden(X,TPTR,SORH,SCAL,N,'wname'). But do you have more information about the arguments? Because, I don&#8217;t even know what do they represent. I try to compare the noise suppression with different parameters, but it is quite hard to make the difference. Do you know for example, which wname I should use with an audio signal?
> > > - My second question is, is there other methods to suppress the noise? Like using a low-pass filter to suppress high frequencies that are in certain case not useful.
> > >
> > > Thank you very much for your help.
> > > +++
> >
> > Hi, have you looked at the examples in the documentation for the wavelet denoising?
> >
> > >>doc wden
> >
> > There's really not a best answer to how to choose most of the input arguments in wden(). The answer is: it depends. I would suggest you go through the demo
> > denoisingsignalsdemo.m which addresses some of this with examples. You can access that demo by typing
> > >>edit denoisingsignalsdemo.m
> > at the command prompt or by Start -> Toolboxes -> Wavelet -> Demos
> >
> > I think the answer to your second question is yes. In many cases simply filtering the signal may yield a very satisfactory denoised version. Again, it depends on a number of factors. If you can give people a little better idea of the kind of signals you are working on and your goals, someone could give you more specific help.
> >
> > Wayne
>
> Thank you Wayne for your answer.
>
> I read it carefully and it was really useful! I am trying to compare different ways of filtering noise in audio signal.
> About "simple filtering" of the signal, what should I filter? only really high frequencies (what is the limit?), or are there other things to do?
>
> Thanks!

Hi,
Again, it depends on what you are defining as noise. Often what people mean by "noise" are high frequency oscillations. In that case, a simple lowpass filter (Fourier filtering) may give you the desired result. However, if the "signal" exhibits certain high frequency features (small scale) mixed in with low frequency (large scale) features, and you want to preserve both, then lowpass filtering will smooth out the small scale features. In that case, wavelet denoising might work better.

You have told us that you have an audio signal, but is your audio signal speech? If so, is the goal to improve the intelligibility of the speech signal? If that is the case, what is the sampling frequency? We minimally need to know something about the spectral support of the "signal" and the sampling frequency to suggest an appropriate filter.

There are a wide range of "audio" signals with very different frequency content.

I think you will need to provide these types of details before anyone can offer you a more concrete way to proceed.

Hope that helps,
Wayne
 | 
Pages: 1
Prev: Hough Transform
Next: synchonous machine