[Speex-dev] Encoding problem

[Jean-Marc Valin]

Your problem (silence being inserted) is due to the lookahead in the
codec. You need to chop the silence both at the beginning and at the
end. What I suggest is to look at how it is done in speexenc and
speexdec. When you use the new version of these utilities, you will see
that they remove all the silence regions.

	Jean-Marc

Le vendredi 05 novembre 2004 à 21:15 +0500, Raheel Qutab a écrit :
> I am using speex for encoding and decoding speech files in a speech
> synthesis program. I am using concatenative approach for speech
> synthesis using diphones (small speech chunks) as basic unit.
> 
>  
> 
> Now in my application, the order of diphones to be joined after
> decoding is determined at runtime. 
> 
> Therefore I have to encode each diphone independent of others.
> 
>  
> 
> I find out that when I encode diphones  in order of their
> concatenation with out reinitializing the encoder for each diphone
> 
> enc_state = speex_encoder_init(&speex_wb_mode); 
> 
> After decoding and concatenation it worked fine. 
> 
> If the order is not maintained concatenated wave output is not as
> clear as when maintaining the order. The problem is traced at the
> joining point of those diphones.
> 
>  
> 
> However when I encode diphones after reinitializing encoder for each
> diphone, there is a small silence region at joining point of every two
> adjacent diphones.  
> 
>  
> 
> It seemed that some sort of smoothening is applied at the encoding
> time between two frames
> 
>  
> 
> How can I get rid of that silence region at joining point of diphones.
> 
>  
> 
> I need help for the above mentioned problem.   
> 
>  
> 
> Regards
> 
> Raheel Qutab
> 
>  
> 
>  
> 
>  
> 
> 
> _______________________________________________
> Speex-dev mailing list
> Speex-dev at xiph.org
> http://lists.xiph.org/mailman/listinfo/speex-dev