[Speex-dev] Downsampling

Thorvald Natvig speex at natvig.com
Tue Jun 7 14:00:38 PDT 2005

> For transforming stereo to mono, averaging is fine and that's what
> everybody does. For sampling rate conversion, it's another matter (too
> long for this email) and you should read a bit about it a perhaps grab a
> library that does that.

Working on it :) As far as I can see, as long as I do downsampling by an 
integer factor, the FIR-filter decimator should do the job with a minimum 
of distortion and aliasing.

On a sidenote, people might want to be aware that many soundcards work 
internally at some specified frequency (usually 48 khz) and if you sample 
at a lower frequency, it simply picks every Nth sample to achieve this. So 
you will often get better signal quality if you sample at the internal 
frequency and then downsample in software.

> As for echo cancellation, it will be less complex (and as good) on a
> (cleanly) down-sampled signal (and certainly not on stereo).

That's good to know. Thanks :)

More information about the Speex-dev mailing list