I have a feeling this question has a lot of research into it, but I can't find any relevant results.
I'm trying to compare the similarity of audio
Here, I have 2 virtually identical samples; however, sample B (blue) is missing a fraction of its tail (making it a subsection of sample A). The samples have had their lengths normalised, which is why they go out of phase.
This de-phasing (due to B being a subsection of A) ruins my analysis of the difference. This problem seems identical to trying to implement image segmentation.
I learnt about this problem on an AI course, on the topic of image segmentation. The method covered there, to find all occurrences, was to brute force search and classify all possible subsections.
I could do something similar (brute force search all subsections of the waveform, and perform the difference analysis), but this seems incredibly expensive. Especially as I want to perform this bidirectionally (either could be a subsection of the other).
Are there any established methods for tackling this?