[gbh444g]

Hi HN! In my previous post, https://news.ycombinator.com/item?id=25037784, I presented so-called ACF images, where the auto-correlation function of a sound waveform is drawn in polar coordinates. It turned out that ACF images capture a good deal of sound symmetry.

This post is a natural extension of that idea. I've been thinking how to introduce colors into ACF images. ACF splits a waveform into a set of pure cosine waves and aligns them together by removing the phase. The idea is to color each cosine wave with the note it corresponds to, so when these waves are aligned, not only the amplitudes of the waves add up, but also their colors.

[nitrogen]

Are the overall amplitudes normalized? Some images appear brighter or darker than others, so I wonder if comparing them would be easier if there was some kind of perceptual brightness normalization step.

[gbh444g]

Amplitudes are normalized locally, using something like a moving average. Global normalization rarely works because most sounds have high dynamic loudness range.