Chinese accent detection research based on RASTA - PLP algorithm

Abstract

Accent is a critical important component of spoken communication, which plays a very important role in spoken communication. In this paper, we conduct accent by using RASTA - PLP algorithm to extract short-time spectrum features of each speech segment based on sub-segment splicing information. We build short-time spectrum feature sets based on RASTA - PLP algorithm. And we choose NaiveBayes classifier to model the feature sets. NaiveBayes is to choose the class with maximum posteriori probability as the object's class. This classification method makes full use of the related phonetic features of speech segment. Based on short-time spectrum of RASTA - PLP feature sets respectively achieve 80.8% accent detection accuracy on ASCCD and on ASCCD (NOISEX92-white). The experimental results indicate that based on sub-segment splicing feature structured method of RASTA - PLP can be used in Chinese accent detection study. RASTA-PLP algorithm is robust on ASSCD and on ASSCD (NOISEX92-white).