> Hi Gabriel,
> 
> >  It uses peak detection. If a freq amplitude is higher by a threshold than
> > its neighbours, then it's considered as tonal. There is no delay like in
> > gpsycho, but if several tones are close enough, it will miss them (could it
> > be the case with Fatboy?).
> 
> I think main problem of nspsytune's tonality estimation method is that frequency
> resolution of 1024 point FFT is not enough. But, simply using longer FFT degrades
> time resolution, and it may cause another problem. So, I'm implementing new
> tonality estimation algorithm which uses long FIR filter instead of FFT. Since
> I have not so much spare time currently, it's implementation is proceeding only
> little by little.
> 
>   Naoki Shibata   e-mail : [EMAIL PROTECTED]
> 

Hi Naoki,

I dont understand this claim, because for any FIR filter, you 
can construct a very good approximation using the FFT.  The longer
the FFT, the better the approximation.  And for a long FIR filter,
the fft will be more efficient.  

Mark

_______________________________________________
mp3encoder mailing list
[EMAIL PROTECTED]
http://minnie.tuhs.org/mailman/listinfo/mp3encoder

Reply via email to