Re: (#484) Speakfreely patches

> * I have written resampling routines to change sampling frequencies. On _some_
> hardware this has made huge improvements to the audio quality. Basically, the
> sampling and playback is done at 44.1kHz and the conversion is made to 8kHz in
> software. I have used a proper polyphase filter implementation to prevent
> artifacts and aliasing. Some soundcards seem to do a very poor job at 8kHz
> sampling, and this resampling has given much better results. This was working
> on both windows and unix in 7.1. I've ported to 7.6 on unix, but haven't
> tested it - it's not useful to me personally at the moment.

Could be interesting for handing the voice data over to a third-party
process. A possibility to output to not only the soundcard but also eg.
stdout could be beneficial. (I could take a look at it sometimes.)

(Another interesting extension I could play with is a simple control
interface, possibly a listening UDP socket, that could facilitate simple
interprocess communication - switching transmitting on/off, later maybe
establishing connection to specified IP - so allowing running sfmike as a
daemon, etc.)

> I also plan the following:
> * Add the Speex codec. That's for the better bitrate/quality tradeoff, and
> because it has algorithms to extrapolate through lost packets. I'm thinking
> both unix and windows versions for this.
>
> * I may have a try at echo-cancellation and comfort noise - especially as
> speex has algorithms for the former.

That's VERY interesting!

> If anyone is interested in these patches, or to try merging them, let me know.
> It seems that the sourceforge projects aren't really going yet either.

I am, it could be interesting at least to look. I could try to merge them
with my security-related patches.

You may also like to check this: http://www.sci.fi/~mjkoskin/ There are
some programs there for denoising audio; they are for wav files, but they
maybe could be rewritten to work on realtime audio. Could offer further
quality improvement, especially with lower-end hardware. (My experiences
with audio processing are somewhere between zero and negligible so can't
be of much help here. :( )


                      * * *

To unsubscribe from this mailing list, send E-mail containing
the word "unsubscribe" in the message body (*not* as the
Subject) to [email protected]