Performance Evaluation of Automatic Mandarin Pronunciations Assessments Based on Fuzzy Integration
Aiming at performance evaluation of Automatic Mandarin Pronunciations Assessments (AMPA) which are applications of perception and acoustics, a general idea based on fuzzy logical is developed: the evaluation algorithm and their indices must be chosen; the comparative information of various assessment algorithms being referred is presented. Then a performance evaluation model is formed based on Sugeno integration. In this paper, four algorithms for AMPA are evaluated: goodness of pronunciation (GOP), confidence based phoneme (CBP) assessment, prosodic features evaluation (PFA) and latent pronunciation analysis (LPA). The experiments simulations are carried out by test data including three typical errors for foreign learners and results show the best performance made by CBP.
automatic Mandarin pronunciation estimating performance evaluation fuzzy function Sugeno integration pronunciation errors detection
Ru Zhang Jiqing Han Jiu Lou
Computer Science and Technology, Harbin Institute of Technology Harbin, China
国际会议
成都
英文
1568-1572
2010-12-17（万方平台首次上网日期，不代表论文的发表时间）