Package HRIRs with additional samplerates

[orbisvicis]

48000Hz is pretty common, and matlab doesn't work on linux (perhaps octave would run prepare_hrirs_kemar.m, untested).

I'm not even sure if ssr resamples everything to 44.1Khz internally anyway. TBH I think I've hit every possible bug, so haven't been able to generate any output.

[mgeier]

This is related to #16.

Currently, the SSR resamples all the source signals, but it doesn't resample the impulse responses.
You can change the sample rate in the JACK settings, the SSR uses the samplerate that JACK uses.

You can use sox to convert the IR files:

sox old_ir.wav -r 48000 new_ir.wav

We might add automatic IR samplerate conversion in the future, stay tuned ...

[orbisvicis]

Isn't this sort of upsampling a bad idea, leading to all sorts of glitches and sub-par quality? I don't know about SSR, but VoxForge/CMUSphinx (speech recognition) both warn against upsampling.

So SSR works in 44.1Khz internally?

[mgeier]

Isn't this sort of upsampling a bad idea, leading to all sorts of glitches and sub-par quality?

Sure, it will never be perfect. The artifacts should be tiny, though, probably this will never be audible in a real-life situation.

However, I don't know for sure, that's why I didn't implement automatic resampling (yet).

I don't know about SSR, but VoxForge/CMUSphinx (speech recognition) both warn against upsampling.

You should avoid it if you can, I agree with that.
But what else should you do if the HRIR measurements were done in a different sampling rate?

I think upsampling is less problematic than downsampling, because in latter you might get aliasing artifacts (in addition to the artifacts you'll get anyway).

So SSR works in 44.1Khz internally?

The SSR uses whatever JACK uses. So you can choose which sample rate to use when starting jackd.