Here's a demo of the stabilized audio image extraction used in CARFAC
(Lyon). One of my students (Iver Jordal) approaches this technique in
an AI based audio processing application. It struck me that SAIs might
be good for a robust pitch tracking method. It could probably be
approached in many ways, but looking at the horizontal spacing of the
frequency "bands" in the visualization of the SAI seems like it could
be used as the basis for pitch estimation.
https://carfac.azurewebsites.net/cpp/carfac.html