evictor
commented
1 year ago I probably can't help you directly, but you should post a diff-style screenshot with A on top half and B on bottom half, graphs with same X scale where X is time (i.e. A would be wider and B would be beneath it aligned somewhere along X by time) and Y is whatever you need to do to convey hash (e.g. int representation of the hash). That way, one could see quite easily and quickly the apparent qualitative differences, like for example what approx. % of hashes were correct, where were the correct ones, etc. and note other potential anomalies that help debug, e.g. all the correct hashes were at the end (or other patterns of that nature).
I will say that B being a subset of A will most likely not have quantized along the same "beat", which is why there is some concept of "overlap" for the windows used to pick out predominant frequencies. I probably botched the terminology, I can't remember what the vars are called in the code.
I have two audio files A and B, B being cut from A i.e B is a subset of A. When I generates hashes for A and B, (A has more hashes than B as B is a subset), not all the hashes of B matches with A, though a lot do. I deally, all the hashes of B should be a subset itself of hashes of A, which is not the case. My major task is alignment. Can you help me out!