Newsgroups: comp.dsp,comp.speech
Path: cantaloupe.srv.cs.cmu.edu!bb3.andrew.cmu.edu!news.sei.cmu.edu!cis.ohio-state.edu!magnus.acs.ohio-state.edu!math.ohio-state.edu!howland.reston.ans.net!newsserver.jvnc.net!newsserver.egr.uri.edu!ramli
From: ramli@ele.uri.edu (Ramli)
Subject: Re: Wigner-Ville distribution
Sender: news@egr.uri.edu (USENET News System)
Organization: URI Department of Electrical Engineering
Message-ID: <CxBL3E.Gn5@egr.uri.edu>
References: <36uihr$1fs@eagle.natinst.com> <36vestINNgt9@dubhe.anu.edu.au> <b.spedding.441.0009CC2E@irl.cri.nz>
Date: Fri, 7 Oct 1994 20:33:14 GMT
Lines: 50
Xref: glinda.oz.cs.cmu.edu comp.dsp:13801 comp.speech:3462

In article <b.spedding.441.0009CC2E@irl.cri.nz> b.spedding@irl.cri.nz writes:
>In article <36vestINNgt9@dubhe.anu.edu.au> pjk@faceng.anu.edu.au (Peter Kootsookos) writes:
>>From: pjk@faceng.anu.edu.au (Peter Kootsookos)
>>Subject: Re: Wigner-Ville distribution
>>Date: 6 Oct 1994 00:03:41 GMT
>
>
>>Has anyone out there found a real use for TFRs (Time-Frequency 
>>Representations)?  
>
>>Other than the short-time Fourier transform, which has been
>>used for many years in the speech analysis/synthesis/processing
>>community, I don't know of any bona-fide application where TFDs
>>outperform more direct techniques.
>
>I've been using them to examine acoustic emission data, they clearly show the 
>arrival of different wave modes.
>
>Bruce
>

I am curious to know if using TFR was the _only_ way with which
you could identify the arrival of different modes for your application.
Is it possible that although the TFR you are using does clearly
show what you are looking for, there might be other conventional techniques
that produce similar results ?

Here are some polemical remarks:

In my (probably limited) experience, I am yet to see a convincing example
which shows that TFR is the _only_ approach that works.  I am of the view
that none of the TFRs have _significantly_ outperformed the STFT (or
spectrogram).  For example, if you take a _challenging_ real-world signal
like voiced-speech, which is highly nonstationary and has many closely
spaced components, none of the existings TFRs have significantly
outperformed the STFT or have revealed hitherto unknown features.

I am as yet unconvinced, although I am willing to change my opinion if
someone shows me a compelling example.

ramli.

PS:  The newsfeed here is pretty erratic; if you feel that these remarks
     cannot go unchallenged and want to throw in your $0.02's worth,
     I'd appreciate if you can mail me a copy just to make sure I get it.

PPS: I have added comp.speech to the list of newsgroups since I have
     mentioned about speech.


