Three techniques for improving automatic synchronization between music and lyrics: Fricative detection, filler model, and novel feature vectors for vocal activity detection.

Hiromasa Fujihara, Masataka Goto

Browse the full ICASSP paper archive.