Sound Classification with Core ML 3

I am interested in knowing what specific features that the machine model extracts and proceeds to use for sound classification?