Hello Jean-Marc. A few weeks ago you wrote on this list:> Samples should never reach saturation (+-32767) and should generally > be less than 16384 to have good results.I would like to use Speex in 8 kbps narrowband mode to compress an audio signal (speech) with several peaks up to +-24000. Would it make sense to reduce the volume by 3 dB before feeding it to Speex and increasing it again afterwards? I did a few tests without noticing much difference. Is it therefore sufficient if 90% of the samples are in the +-16384 range? Thank you, Daniel
I think it should be fine. Just test and see if you're not sure. Jean-Marc Daniel Schmidt a ?crit :> Hello Jean-Marc. > > A few weeks ago you wrote on this list: > >> Samples should never reach saturation (+-32767) and should generally >> be less than 16384 to have good results. > > I would like to use Speex in 8 kbps narrowband mode to compress an > audio signal (speech) with several peaks up to +-24000. Would it make > sense to reduce the volume by 3 dB before feeding it to Speex and > increasing it again afterwards? I did a few tests without noticing > much difference. Is it therefore sufficient if 90% of the samples are > in the +-16384 range? > > Thank you, > Daniel > _______________________________________________ > Speex-dev mailing list > Speex-dev@xiph.org > http://lists.xiph.org/mailman/listinfo/speex-dev > >