[openbeos] Re: Resampling algorithms for audio mixer

2007/11/28, cyanh256@xxxxxxxxxxxx <cyanh256@xxxxxxxxxxxx>:
> > What about "interpolating remaining one sample behind the input
> > buffer" (so that you cannot have to interpolate values beyond  the
> > innput buffer) and passing the last 3 or 4 samples of the previous
> > input buffer through a struct, as you suggested, so that the values
> > you now have to inteprolate before the input buffer are well defined?
>
> Exactly; that's pretty much what I was proposing. Preserving
> state should clear up any potential concerns about sound quality
> with very small buffer sizes (as an extreme example, one-sample
> buffers!).
>
> It's been a while since I wrote any resampling code, but the
> last one I wrote worked like this: (this is from memory, not
> actual cut+paste code, so it probably won't work, and certainly
> isn't optimized!)
>
> typedef struct RESAMPLER {
>     int32 cur, prev;
>     uint32 pos; /* Current fractional position */
> } RESAMPLER;
>
> /*
>  * Produce an output buffer at the desired sample rate,
>  * reading the input stream as required.
>  * 'out_rate' is the output sample rate relative to the input sample
>  * rate in 16:16 fixed point format. e.g., with a 44100Hz output
>  * rate and an 88200Hz input rate, 'out_rate' must be 32768.
>  */
> void resample(
>     RESAMPLER *state,
>     int16 *out_buf,
>     size_t out_len,
>     uint32 out_rate
>     ) {
>     for( ; out_len-- ; ) {
>         *out_buf++ = (int16) (( state->cur * state->pos + state->prev *
>  (~state->pos) ) >> 16);
>         state->pos += out_rate;
>         while( state->pos & 0xffff0000 ) {
>             /* The position has overflowed into the next sample --
>                read a new sample */
>             state->pos -= 0x10000;
>             state->prev = state->cur;
>             state->cur = READ_SAMPLE(); /* This returns an int16 */
>         }
>     }
> }
>
> The architecture is pretty different given that it uses an 'out_rate'
> value
> rather than using the relative sizes of the two buffers, and reads the
> input stream internally rather than having a separate input buffer.
> Though the idea of state preservation doesn't depend on that, of
> course.

Agreed, then.

I'm waiting for you guys to tell me what to do exactly
(library/integration into other libraries/just code/whatever).

Anyway I have to warn you that I can work on it only in my "spare
spare time", which means that I already work on a project of mine in
my spare time (it can also interest you a lot, and I do hope so -
maybe we can talk about it later), so don't expect it for the next
week, ok? ;-)

Stefano

Other related posts: