During hardware and network impairment testing, I sometimes just hold down a DTMF key to keep the audio busy. I have found that if I restart the receiver while the transmitter is still encoding DTMF, the receiver plays out a different tone, until the DTMF key goes off an on again. Also, today I captured some Speex frames during a tone (but not including the start of the tone). When I later fed these into the decoder, I got mostly silence with some noise. Is this behavior expected? I suspect it is, and I understand that this is not a real-world situation. This testing was done with 1.2beta1, 8KHz, 8kbps, TI C55 DSP, no preprocessor or echo canceler. Thanks, Jim -------------- next part -------------- An HTML attachment was scrubbed... URL: http://lists.xiph.org/pipermail/speex-dev/attachments/20061108/a9a02e42/attachment.htm
Jim Crichton wrote:> During hardware and network impairment testing, I sometimes just hold > down a DTMF key to keep the audio busy. I have found that if I restart > the receiver while the transmitter is still encoding DTMF, the receiver > plays out a different tone, until the DTMF key goes off an on again. > > Also, today I captured some Speex frames during a tone (but not > including the start of the tone). When I later fed these into the > decoder, I got mostly silence with some noise. > > Is this behavior expected? I suspect it is, and I understand that this > is not a real-world situation.It's a bit funny, but I'm not too surprised. what happens is that with these signals, the prediction gain is so big that there's nearly no need for and "innovation" (fixed codebook) signal. When you reset the decoder, the states become out of sync until the prediction gains drop. You can always play with the gain control if you want to limit the effect. Right now, I've got if (cumul_gain > 262144) max_gain = 31; in ltp.c around line 360. You can reduce the value (especially the first one) and see what it does. Most likely it'll improve in the situation you're describing at the price of a slight deterioration in the normal case. Jean-Marc
Possibly Parallel Threads
- Fwd: Re: Fixed Point on wideband-mode: Single Frame loss on 2000 Hz sine causes "freak off"
- Speex echo canceller on TI C55 DSP
- AEC on a TI C6x - has no effect
- Fwd: Re: Fixed Point on wideband-mode: Single Frame loss on 2000 Hz sine causes "freak off"
- Speex echo canceller on TI C55 DSP