General Discussion

Nested
extracting features after forced alignment
User: chutzzz
Date: 8/24/2012 6:53 am
Views: 3686
Rating: 9

Is it possible to extract features of each phone/phoneme after forced alignment?

 

I was already able to do forced alignment, sadly the features extracted are only those of the ENTIRE audio file. What I wish to do is to extract the features of each alignment.

 

BTW, the features I am talking about are 12th order MFCCs, energy and first and second derivatives.

 

Please help. Many Thanks.

Ian

 

--- (Edited on 8/24/2012 6:53 am [GMT-0500] by ) ---

Re: extracting features after forced alignment
User: nsh
Date: 8/27/2012 2:33 pm
Views: 1703
Rating: 11

> Is it possible to extract features of each phone/phoneme after forced alignment?


Yes, you can use HCopy with start and end times returned by aligner

> What I wish to do is to extract the features of each alignment.


You can easily do that with a simple script

 

--- (Edited on 8/27/2012 23:33 [GMT+0400] by nsh) ---

PreviousNext