Automatic Transcription of Vocal Quartets

Abstract

This work presents a probabilistic latent component analysis (PLCA) method applied to automatic music transcription of a cappella performances of vocal quartets. A variable-Q transform (VQT) representation of the audio spectrogram is factorised with the help of a 6-dimensional tensor. Preliminary experiments have shown promising music transcription results when applied to audio recordings of Bach Chorales and Barbershop music.