Skipped content of type multipart/alternative-------------- next part -------------- A non-text attachment was scrubbed... Name: output.wav Type: audio/x-wav Size: 259244 bytes Desc: not available Url : http://lists.xiph.org/pipermail/speex-dev/attachments/20061127/546b3a06/output-0001.wav -------------- next part -------------- A non-text attachment was scrubbed... Name: input.inp Type: chemical/x-gamess-input Size: 259200 bytes Desc: not available Url : http://lists.xiph.org/pipermail/speex-dev/attachments/20061127/546b3a06/input-0001.bin
Hi Lianghu, I was indeed able to reproduce the problem. I'll try and fix that soon. About the earlier fixed-point problem with wideband on a very high volume (and heavily clipped) sample, it should be much better now. I'm also in the process on cutting down on wideband RAM requirement. The version in svn already requires much less RAM than 1.2beta1 and there's still a lot of room for improvement in the encoder (decoder's now very small). Jean-Marc lianghu xu a ?crit :> Hi All, > > Please consider the WB case with following parameter, > > INPUT : input.inp > MODE: SPEEX_Wideband > RATE: 12.800kbps > DTX: DTX_ENABLED > VAD: VAD_ENABLED > > > The fixed-point code in current SVN is used. > The output after wideband coding and wideband decoding is also attached. > > I think the ouput is unacceptable in its non-speech part. > > > Please check it. > > Regards, > > Lianghu > > > ------------------------------------------------------------------------ > > _______________________________________________ > Speex-dev mailing list > Speex-dev@xiph.org > http://lists.xiph.org/mailman/listinfo/speex-dev
> The fixed-point code in current SVN is used. > The output after wideband coding and wideband decoding is also attached. > > I think the ouput is unacceptable in its non-speech part.Should be fixed in SVN now. Can you confirm? Jean-Marc P.S. The mailing list sometimes has a latency of several days, sorry for the inconvenience.
Hi, I've just try the SVN and it's OK! I think the critical update is in cb_search.c line 583, tmp[i]=SHL32(EXTEND32(target[i]),8); Could you give an explanation to the change? Thank you, Lianghu On 12/2/06, Jean-Marc Valin <jean-marc.valin@usherbrooke.ca> wrote:> > > The fixed-point code in current SVN is used. > > The output after wideband coding and wideband decoding is also attached. > > > > I think the ouput is unacceptable in its non-speech part. > > Should be fixed in SVN now. Can you confirm? > > Jean-Marc > > P.S. The mailing list sometimes has a latency of several days, sorry for > the inconvenience. >-------------- next part -------------- An HTML attachment was scrubbed... URL: http://lists.xiph.org/pipermail/speex-dev/attachments/20061204/e4f79a7c/attachment.htm