OnsetSlice Metrics 0 & 1 don't compare frames?

[tedmoore]

Perhaps I'm missing something here, but it looks like these two metrics in OnsetSlice don't compare frames, as they don't take in (or use) a Ref to prev as the others do (some others also use a prevprev when the measurement requires it). It looks like they just compute the value for a single frame and return that.

Thoughts:

• Are these helper functions that compute the needed feature for a single frame and the comparison happens elsewhere? I couldn't find where.
• Are these metrics not intended to compare magnitudes of different frames, but are meant to compute a single feature value for each frame and see if each frame is above the raw threshold?

https://github.com/flucoma/flucoma-core/blob/e3cadeb8382e9e473f3b5f94a2f3a0c93fa082a7/include/flucoma/algorithms/util/OnsetDetectionFuncs.hpp#L40-L50

[tedmoore]

The docs suggest that frames are being compared.

[tremblap]

They have been coded from the SC paper that implements them, so it might just be a documentation issue. They do indeed seem to be absolute and not relative.

[tedmoore]

Do you have a link to the paper? Just curious to see it.

[tremblap]

should be in the help file, let me see

[tremblap]

referred to in the learn platform, and it is also in the SC source code

D. Stowell and M. D. Plumbley. Adaptive whitening for improved real-time audio onset detection. Proceedings of the International Computer Music Conference (ICMC2007), Copenhagen, Denmark, August 2007. See http://c4dm.eecs.qmul.ac.uk/papers/2007/StowellPlumbley07-icmc.pdf

[tedmoore]

So should the conclusion be to change the documentation to match the algorithm?

[tremblap]

indeed. I'm moving this to docs