Singing Power Ratio (SPR)

i wouldnt use cross~, you normally use FFT for that kind of thing. the higher steepness will beat the less exact split frequency.

you should also be aware that the whole idea only works good if you use the right material.

i.e. cut off the attack from voices and instruments, normalize all files to zero, avoid intereferences, noise, hissing, popping, then use several takes over several notes from the materials to be compared, stuff like that.