> >
> >True, depends on the application. If detection-related then I think
> >you want to normalize. Maybe DWT can report back as to what worked
> >best. :)
> >
> >KP
> >
>
> Ken and illywhacker, thanks again for the valuable explanations.
>
> Here you'll find the screenshot:
> http://img145.imageshack.us/img145/4331/kenwf5.jpg
>
> As you can see the original Ken's method (with additional 1\log step)
> provides more accurate boundaries for the /s/ in the second word.
>
> On the other hand, spectral entropy (w\o normalization) shows the same
> results with standard entropy (the method based on wavelet coefficients
> energy).
>
> By the way, Ken, is it your original idea of using additional
> normalization or is it taken from the paper\book?
>
>True, depends on the application. If detection-related then I think
>you want to normalize. Maybe DWT can report back as to what worked
>best. :)
>
>KP
>
Ken and illywhacker, thanks again for the valuable explanations.
Here you'll find the screenshot:
http://img145.imageshack.us/img145/4331/kenwf5.jpg
As you can see the original Ken's method (with additional 1\log step)
provides more accurate boundaries for the /s/ in the second word.
On the other hand, spectral entropy (w\o normalization) shows the same
results with standard entropy (the method based on wavelet coefficients
energy).
By the way, Ken, is it your original idea of using additional
normalization or is it taken from the paper\book?
Reply by KP●February 18, 20092009-02-18
On Feb 18, 4:13�am, illywhacker <illywac...@gmail.com> wrote:
> On Feb 17, 11:40�pm, KP <ken.pra...@gmail.com> wrote:
>
> > On Feb 17, 1:08�am, illywhacker <illywac...@gmail.com> wrote:
>
> > > On Feb 17, 12:17�am, Ken Prager <pra...@ieee.org> wrote:
> > > > Additional normalization will make Ei range between 0 and 1.
>
> > > Is that good?
>
> > Yes--makes it easier to compare across multiple subbands.
>
> I guess, although it is not clear that this is necessarily the right
> way to compare, is it?
>
> illywhacker;
True, depends on the application. If detection-related then I think
you want to normalize. Maybe DWT can report back as to what worked
best. :)
KP
Reply by illywhacker●February 18, 20092009-02-18
On Feb 17, 11:40�pm, KP <ken.pra...@gmail.com> wrote:
> On Feb 17, 1:08�am, illywhacker <illywac...@gmail.com> wrote:
>
> > On Feb 17, 12:17�am, Ken Prager <pra...@ieee.org> wrote:
> > > Additional normalization will make Ei range between 0 and 1.
>
> > Is that good?
>
> Yes--makes it easier to compare across multiple subbands.
I guess, although it is not clear that this is necessarily the right
way to compare, is it?
illywhacker;
Reply by KP●February 17, 20092009-02-17
On Feb 17, 1:08�am, illywhacker <illywac...@gmail.com> wrote:
> On Feb 17, 12:17�am, Ken Prager <pra...@ieee.org> wrote:
> > Additional normalization will make Ei range between 0 and 1.
>
> Is that good?
Yes--makes it easier to compare across multiple subbands.
Ken P.
Reply by DWT●February 17, 20092009-02-17
>On Feb 17, 10:16=A0am, "DWT" <zwfilte...@yahoo.com> wrote:
>> >The situation is exactly as Ken expresses it. There are two different
>> >notions: the entropy of the normalized wavelet coefficient energies,
>> >and the entropy of the normalized wavelet Fourier coefficient
>> >energies. These are not the same. Indeed you can choose any basis you
>> >like and you will have a different answer.
>>
>> >What are you trying to do with this?
>>
>> >illywhacker;
>>
>> Voice activity detection.
>
>Hasn't this kind of technique been much used already in that area? Are
>you trying to do something new, or use something already done?
>
>illywhacker;
>
You know that the best way to do something new is to understand how the
existing methods work.
I'm just trying to build simple VAD based on the renowned features now.
Then I'll try to do something new. It would be great, if you can give me
any idea\direction on new VAD speech features, too!
Reply by illywhacker●February 17, 20092009-02-17
On Feb 17, 10:16�am, "DWT" <zwfilte...@yahoo.com> wrote:
> >The situation is exactly as Ken expresses it. There are two different
> >notions: the entropy of the normalized wavelet coefficient energies,
> >and the entropy of the normalized wavelet Fourier coefficient
> >energies. These are not the same. Indeed you can choose any basis you
> >like and you will have a different answer.
>
> >What are you trying to do with this?
>
> >illywhacker;
>
> Voice activity detection.
Hasn't this kind of technique been much used already in that area? Are
you trying to do something new, or use something already done?
illywhacker;
Reply by DWT●February 17, 20092009-02-17
>If I follow your math, you've now calculated subband entropy instead of
>subband spectral entropy. It might work, depending on what you are
>planning to use it for. You might calculate both for a few different
>signals and compare how they track. One method may be better than the
>other for your application.
Yes, Ken. It's the energy entropy. I understand the difference. I'm just
looking for reliable (and quick) alternative for a energy feature for VAD.
Reply by DWT●February 17, 20092009-02-17
>The situation is exactly as Ken expresses it. There are two different
>notions: the entropy of the normalized wavelet coefficient energies,
>and the entropy of the normalized wavelet Fourier coefficient
>energies. These are not the same. Indeed you can choose any basis you
>like and you will have a different answer.
>
>What are you trying to do with this?
>
>illywhacker;
>
Voice activity detection.
Reply by illywhacker●February 17, 20092009-02-17
On Feb 16, 9:10�pm, "DWT" <zwfilte...@yahoo.com> wrote:
> >> My warmest apologies.
>
> >> 1) Calculating subband energy:
> >> E=3Dsum(|Wij|.^2),
> >> where Wij - k-th coefficients on the current decomposition level i.
>
> >> 2) Energy-entropy:
> >> Entr=3D-sum(E*log(E))
>
> >> I know that it's not the spectral entropy.
>
> >> I'll use Ken's method as well, but I'm also looking for the quick
> methods=
> >.
>
> >Sum over what?
>
> >illywhacker;
>
> Oops, I mistyped. Here's the revised algo:
> W(i,j) - wavelet coefficients (WC), where i - decomposition level.
>
> On each level 'i':
> 1) Getting each coefficient's energy:
> E(i,j)=W(i,j).^2;
>
> 2) Normalize this energy:
> NE(i,j)=E(i,j)./sum(E(i,j)); [sum over j]
>
> 3) Normalized Shannon Entropy:
> Entr(i)=-sum(NE(i,j)*log(E(i,j))) [sum over j]
>
> Is this a viable method for entropy calculation?
The situation is exactly as Ken expresses it. There are two different
notions: the entropy of the normalized wavelet coefficient energies,
and the entropy of the normalized wavelet Fourier coefficient
energies. These are not the same. Indeed you can choose any basis you
like and you will have a different answer.
What are you trying to do with this?
illywhacker;