Research Seminars

Probabilistic Information Integration for Music Scene Analysis
Kunio Kashino, NTT Communication Science Laboratories, Japan
Wednesday 4th December 2002

This talk focuses on probabilistic information integration methods for music scene analysis. The analysis means describing notes and chords from an audio signal of music. The problem is viewed in terms of inference to find the most plausible interpretation of the input signal among multiple hypotheses. Source separation examples based on the recognition results will also be presented.