[hts-users:01411] Re: About Delta, acceleration and HMGenS window configuration

["Heiga ZEN (Byung Ha CHUN)" <zen@xxxxxxxxxxxxxxx>]

Hi,

Qian Xiaojun wrote (2008/05/27 14:45):

>     I'm writing to inquire why the delta and delta delta feature are 
> calculated in this way, as noticed in the window coefficient 
> configuration for HMGenS:
>     Assume the static feature is x( t );     
>     1. delta = -0.5*x( t-1 ) + 0*x( t ) + 0.5*x( t+1 ), intuitively
>     2. deltadelta = 1*x( t-1 ) + (-2)*x( t ) + 1*x( t+1 ), how is this 
> derived?   {[x( t+1 ) - x( t )]/1 - [x( t ) - x( t-1 )]/1}/1 ?

No, they are derived from the second order regression analysis using x(t+1), x(t), and x(t-1).
  
>     Another question is that I use the regression formula on HTK book 
> (5.16) to calculate the delta feature, and delta delta feature is 
> obtained by applying the formula again on delta. Does HMGenS config file 
> support setting the coefficients on a larger window scope, e.g. 5 or 7 
> rather than 3 ?

Yes.

Regards,

Heiga ZEN (Byung Ha CHUN)

--
------------------------------------------------
Heiga ZEN     (in Japanese pronunciation)
Byung Ha CHUN (in Korean pronunciation)

Department of Computer Science and Engineering
Nagoya Institute of Technology
Gokiso-cho, Showa-ku, Nagoya 466-8555 Japan

http://www.sp.nitech.ac.jp/~zen
------------------------------------------------