Hello Jean-Marc. First of all, thank you for your awesome work on speex! In the current version in SVN the wideband encoder/decoder doesn't correctly pass the parameter of SPEEX_SET_SUBMODE_ENCODING to the underlying narrowband codec. My patch fixes this. Then I have a question regarding the EPIC_48K mode. Should the perceptual enhancer work with this mode? At the moment it is enabled by default but only produces a very loud noise. Finally I would like to ask what your near future plans for the (fixed-point) AGC are. I want to add a speaker phone mode to a PDA based VoIP application and AGC would be very useful here. I'm no DSP expert but if I can help you somehow with development or testing, please let me know. Thank you Daniel -------------- next part -------------- A non-text attachment was scrubbed... Name: sb_celp.patch Type: application/octet-stream Size: 905 bytes Desc: not available Url : http://lists.xiph.org/pipermail/speex-dev/attachments/20061209/f67abfc1/sb_celp.obj
> In the current version in SVN the wideband encoder/decoder doesn't > correctly pass the parameter of SPEEX_SET_SUBMODE_ENCODING to the > underlying narrowband codec. My patch fixes this.Thanks. Strange that this didn't cause problems before.> Then I have a question regarding the EPIC_48K mode. Should the > perceptual enhancer work with this mode? At the moment it is enabled > by default but only produces a very loud noise.EPIC_48K wasn't update in a while. I've recently applied a few patches to it, but it needs a few more (the enhancer and the LPC window).> Finally I would like to ask what your near future plans for the > (fixed-point) AGC are. I want to add a speaker phone mode to a PDA > based VoIP application and AGC would be very useful here. I'm no DSP > expert but if I can help you somehow with development or testing, > please let me know.None yet in the short-term, although the long-term plan is for everything to be in fixed-point. Jean-Marc
Hello Jean-Marc, thank you for your reply. I have another question regarding the current AGC algorithm. When it runs several minutes, it doesn't seem to increase the gain anymore but still reduces it immediately (and therefore permanently) after a loud noise, e.g. when the user touches the headset's microphone. Is this behaviour intentional? Regards Daniel