On 18/07/2015, Peter S <peter.schoffhau...@gmail.com> wrote: > I find that what worked well for simple 1-bit signals, doesn't work > well for multi-bit (audio) signals
Here's an example, using a variant of this metric for 8-bit waveforms: http://morpheus.spectralhead.com/img/8bit_saw.png The "entropy" graph at the right shows the estimate with increasing amounts of precision, the width of the bar meaning the result of the estimate. It converges well until some point, at which point, it stops converging to zero, indicating that for general audio signals, this metric will give poor results. What would work better for general audio signals, I'm still trying to find out. > so be careful, lest you lose people's interest Based on the logs of my server, a total of 3 people viewed the results first-order histogram-based entropy estimate. If we assume an average of 100 active readers (I think realistic), then the estimated amount of interest for entropy estimation is 3%. Which is, barely above zero. -P -- dupswapdrop -- the music-dsp mailing list and website: subscription info, FAQ, source code archive, list archive, book reviews, dsp links http://music.columbia.edu/cmc/music-dsp http://music.columbia.edu/mailman/listinfo/music-dsp
