Instead, you apply machine learning (short ML). ML is essentially a combination of some smart math and brute force computation, that lets you extract "useful" information form your data (audio). Useful in quotation marks because that information helps you solve the problem of data matching, it usually does not help to give you an intuitive understanding of what's going on.