Automatic detection and evaluation of the Erhua in Putonghua proficiency test

An automatic detection and evaluation method of the "Erhua" (also called "r-retrofiexion" or "retroflex suffixation") in Putonghua proficiency test (PSC) is proposed. Based on the framework of the computer aided pronunciation evaluation system, this paper makes an in-depth analysis of phonologic rules and acoustic characteristic of the Erhua, solves the detection and evaluation of the Erhua as a typical classification problem. And then more representative acoustic features are selected and a variety of different classification algorithms are used. The results showed that the boosting classification and regression tree (CART) can make full use of the characteristics of the Erhua, and the classification accuracy is 92.41%. Based on the further analysis of the acoustic feature group, we found the formant, pronunciation confidence and duration are the most important clues of the Erhua, and using these clues can effectively realize the automatic detection and evaluation of the Erhua.