A Novel Evaluation Model for Teaching Assessment

It is essential for Mandarin Proficiency Test that mandarin pronunciation is graded by computer system. But it still has a lot of problems need to be solved before it reaches the level of success. This paper proposes a novel method of objective grading pronunciation. The mandarin pronunciation can be decomposed into consonant, vowel and tone. In this paper the pronunciation feature parameters are explored, and Hidden Markov Model (HMM) is apply to assess the pronunciation. Experimental results show that objective grade correlate well with the subjective grade because perception distortions of human auditory have been taken into consideration.