Fundamental frequency estimation and tracking using maximum likelihood harmonic matching and HMMs

A new approach is presented for the estimation and tracking of the fundamental frequency (f0) of pseudoperiodic signals. It is based on a probabilistic model of pseudoperiodic signals that makes it possible to take prior knowledge into account and to include constraints on the evolution of the signal. The resulting method can operate on a large interval of f0 values (typically from 50 to 4000 Hz) and on a great variety of sound signals (speech and music signals).<>