A Multi-Pitch Detecting Method Suitable for CASA
He Zhao · Dianzi xuebao · 2003
This paper puts up a method suitable for multi pitch detecting under overlapping speech signals environment.In this method,wavelet transform is used as filtering analysis part of this pitch detecting model.Besides that,generalized autocorrelation function is used to strengthen pitch information and enhanced summary autocorrelation function is used to weaken redundant information.It is the most important that a pitch probability function is given to predict and tail after each pitch tracking to improve the veracity of pitch detecting.Above mentioned method could be applied to computational auditory scene analysis.From the experiment results provided,we can infer that this method is very useful and efficient.