Single channel polyphonic signal separation based on a novel Multi-F0 estimation method

In this paper, we propose a single channel polyphonic signal separation algorithm based on Bayesian harmonic model and a novel Multi-F0 estimation method. The algorithm estimates Multi-F0 by analyzing the spectrum with a novel Multi-F0 estimation method that consists of two parts. At first, part 1 estimates F0s with no multiple relation, and then the undetected F0s are estimated in part 2. Other audio parameters are extracted by the Bayesian harmonic model which is based on RJMCMC method. The algorithm can be applied to separation of audio mixtures with polyphony of two to four, and experimental results show that the proposed algorithm can separate single channel polyphonic signals effectively.