Automatic Pronunciation Assessment for Mandarin Proficiency Test Based on HMM

Abstract

Objective pronunciation assessment plays a very important role in the Mandarin Proficiency Test. But it still has a long way to go before it reaches the level of success. In this paper, the novel Mandarin objective pronunciation assessment pronunciation of is proposed. The standard of Mandarin pronunciation is divided into six levels. The mandarin pronunciation is divided into consonant, vowel and tone. And the feature parameters are explored, and Bark-SD is selected as consonant assessment parameter. Hidden Markov Model (HMM) is used to assess the pronunciation. The objective assessments correlate well with the subjective assessment because perception distortions of human auditory have been taken into consideration. Experimental results show the method performs better than the exiting methods.