Combined Magnitude Difference Function Based Pitch Tracking Algorithm
Wenhu Wu · Dianzi xuebao · 2006
To reduce the halving and the doubling errors in pitch tracking,an algorithm based on a combined magnitude difference function is proposed in this paper.By analyzing the error distributions in different algorithms based on magnitude difference functions and autocorrelation functions,a combined magnitude difference function(CMDF) is defined,which has the best performance compared with the others.Moreover,a post processing method is proposed which uses the history information to adjust the raw pitch estimated using the CMDF.The post processing method uses only the history information,thus the pitch tracking algorithm has low latency,which reduces the pitch estimation errors that often occur in traditional methods.Experiments show that the pitch tracking algorithm proposed in this paper can achieve a pitch error rate reduction of 13.8% compared with the baseline traditional algorithm.