DSP Robotics Support

In ongoing audio stream - how to kill short-timed artifacts (if the surroudning time is floor) above certain threshold, and still keep such kind of stuff, when it's densier there? I mean - concept.

I'm using threshold detection based on envelope follower, but I'd like to make it better a little bit.

One way, I guess, is to set to max all audio values that are above the threshold, so that envelope follower thinks "this is loud".

But even before that - I guess there could be something more, something else, a sort of short timed compressor or so?

The idea is to hit quickly, release quickly, and keep track on continuity of longer segments, while dampling short pings.

Interesting question!

Do you mean like something to remove crackles on a vinyl record?

I think I can visualise what you mean with the envelope follower and dynamic threshold. I reckon a high pass filter would be needed for detection and a certain delay time (so-called look-ahead) so you could enable an averaging filter when the spike is seen, for a short duration. The delay would be needed to compensate for the detection time and latency of the envelope follower, to get the timing right.

Or…?

Cheers

Spogg

No, audioskip.

If you have audio content with long portions of silence, and short but louder artifacts within that silence - I'd like to skip these artifacts, but keep longer content part.

The audioskip itself is a simple routine, you can trigger (via threshold detection that outputs 0 or 1) normal playback or 100x faster (at the same time silencing audio out); note, that audio playback that goes through silencer - should be delayed a little bit for compensation (due to latency of threshold detection). This part works pretty well here.

Fascinating!

I think I see, so here's my next guess...

Is this to remove "silent" portions in speech, like for an audio book? And you want to not trigger on clicks and bangs from the microphone etc?

Cheers

Spogg

I moved with this to FB and let's continue there.