[Speex-dev] Bug in vbr_analysis

[Jean-Marc Valin]

Hi Aymeric, 

I actually fixed that problem in svn only a couple hours before you sent
your email. You can either update to svn or simply make this:

qual += .3*log(.001+ener/60000.0);

	Jean-Marc
Le mardi 14 février 2006 à 17:20 +0100, Aymeric Moizard a écrit :
> Hi Jean-Marc,
> 
> I used the following code after init of the encoder to
> get the frame size of my RTP packets.
> 
>        char pcm_data[3200];
>        char speex_data[3200];
>        SpeexBits bits;
>        memset(pcm_data, '\0', 3200);
>        memset(speex_data, '\0', 3200);
>        speex_bits_init(&bits);
>        speex_bits_reset(&bits);
>        speex_encode_int(codec_info->enc, (short*)pcm_data, &bits);
>        codec->enc_frame_size = speex_bits_write(&bits, speex_data, frame_size);
> 
> I never run into problems using either speex on linux or windows until one
> used my dll and speex through delphi on windows: he gets division by zero.
> 
> I found that "ener" is 0 around line 253: (version 1.1.11)
> 
>     if (ener<60000)
>     {
>        if (vbr->consec_noise>2)
>           qual-=0.5*(log(3.0 + vbr->consec_noise)-log(3));
>        if (ener<10000&&vbr->consec_noise>2)
>           qual-=0.5*(log(3.0 + vbr->consec_noise)-log(3));
>        if (qual<0)
>           qual=0;
> 
>        qual += .3*log(ener/60000.0);
>     }
> 
> I think this is a side effect of the fact that my pcm_data is
> initialized to '\0'....
> 
> I have no fix except something like: (which is working!)
> 
>    if (ener!=0) qual += .3*log(ener/60000.0);
> 
> I still don't understand why this seems to happen only when
> running inside a delphi program...
> 
> I hope to get a better fix!
> Thanks for all.
> 
> Aymeric
> http://www.antisip.com
> http://sip.antisip.com   --  Free Telephony account.
> 
> 
> _______________________________________________
> Speex-dev mailing list
> Speex-dev at xiph.org
> http://lists.xiph.org/mailman/listinfo/speex-dev
>