The problem is that now I have two files very similar (one for mic, the other for speaker). If I play them together I can hardly say that they are two different file for most of the file length. When I try to cancel the echo, I obtain as output almost the same input. The update_foreground variable is always 0. I don't get the conditions which must be fulfilled in order to have update_foreground=1. I see that there is some energy comparison, but I don't get the physical meaning of all the components. Is it normal to get always 0? If the track are into alignment, are there other tracks' properties to consider in order to get a good echo cancellation? Maybe the audio quality? or the relative volume? Thanks for any help. Marco Pierleoni. -------------- next part -------------- An HTML attachment was scrubbed... URL: http://lists.xiph.org/pipermail/speex-dev/attachments/20091230/dabc69ed/attachment.htm
Maybe you shifted the signal too much. If the mic signal is just one single sample *ahead* of the playback signal, then there's no cancellation possible (because the system is not causal). Be careful with that. Otherwise, if could simply be dropped frames. Jean-Marc On 2009-12-30 07:06, Marco Pierleoni wrote:> The problem is that now I have two files very similar (one for mic, the > other for speaker). > If I play them together I can hardly say that they are two different > file for most of the file length. > > When I try to cancel the echo, I obtain as output almost the same input. > The update_foreground variable is always 0. I don't get the conditions > which must be fulfilled in order to have update_foreground=1. > I see that there is some energy comparison, but I don't get the physical > meaning of all the components. > Is it normal to get always 0? > If the track are into alignment, are there other tracks' properties to > consider in order to get a good echo cancellation? Maybe the audio > quality? or the relative volume? > > Thanks for any help. > > Marco Pierleoni. > > > > _______________________________________________ > Speex-dev mailing list > Speex-dev at xiph.org > http://lists.xiph.org/mailman/listinfo/speex-dev
Thanks Jean-Mark for the answers. I thought about that and I delayed the mic by 50 samples. In this way I get a few update_foreground = 1, but there are not audible differences. Plotting the st->e array, you can see that it is almost null everywhere or with very low values. What it seems weird to me is that if you listen the files they are very similar, so, I think, there should be a way to cancel the echo. Marco 2009/12/30 Jean-Marc Valin <jean-marc.valin at usherbrooke.ca>> Maybe you shifted the signal too much. If the mic signal is just one > single sample *ahead* of the playback signal, then there's no > cancellation possible (because the system is not causal). Be careful > with that. Otherwise, if could simply be dropped frames. > > Jean-Marc > > On 2009-12-30 07:06, Marco Pierleoni wrote: > > The problem is that now I have two files very similar (one for mic, the > > other for speaker). > > If I play them together I can hardly say that they are two different > > file for most of the file length. > > > > When I try to cancel the echo, I obtain as output almost the same input. > > The update_foreground variable is always 0. I don't get the conditions > > which must be fulfilled in order to have update_foreground=1. > > I see that there is some energy comparison, but I don't get the physical > > meaning of all the components. > > Is it normal to get always 0? > > If the track are into alignment, are there other tracks' properties to > > consider in order to get a good echo cancellation? Maybe the audio > > quality? or the relative volume? > > > > Thanks for any help. > > > > Marco Pierleoni. > > > > > > > > _______________________________________________ > > Speex-dev mailing list > > Speex-dev at xiph.org > > http://lists.xiph.org/mailman/listinfo/speex-dev > _______________________________________________ > Speex-dev mailing list > Speex-dev at xiph.org > http://lists.xiph.org/mailman/listinfo/speex-dev >-------------- next part -------------- An HTML attachment was scrubbed... URL: http://lists.xiph.org/pipermail/speex-dev/attachments/20091230/4a82fbcf/attachment-0001.htm