Choice and adaptation of statistical models for single channel singing voice separation

摘要

The problem of singing voice extraction from mono audio recordings, i.e., one microphone separation of voice andudmusic, is studied. The approach is based on a priori probabilistic models for two sources, more precisely on GaussianudMixture Models (GMM). A method for model adaptation to the characteristics of the mixed sources is developed and audcomparative study of different models and estimators is performed. We show that the adaptation of the model of musicudfrom the non-vocal parts of songs yields good results in realistic conditions.