摘要/Abstract

文章通过论证英语口语机考中英语塞音语音识别实现的可能性和实现方法，试图将一种新型的语音识别方法运用在英语口语机考实践中。文章探讨了“Bayes判别准则的2次判别函数法”改进型语音识别方法的最佳方案并通过实验求出了它的最佳参数，从而提出了适合于用“Bayes判别准则的2次判别函数法”改进型语音识别方法进行英语塞音语音识别的语音特征参数变换方法。文章通过对英语塞音语音的识别实验，求证了“Bayes判别准则的2次判别函数法”改进型语音识别方法在英语塞音语音识别过程中具有良好的识别效果。

Through discussion on modified “Quadratic Discriminant Function” method , this paper presents an approach for recognition of English stop consonants in computer-assisted English oral tests based on modified “Quadratic Discriminant Function”. In this paper, we discussed the best way to work out the “Modified Quadratic Discriminant Function” coefficients and presented a kind of phonetic feature coefficients transferring way fit for the use of “Modified Quadratic Discriminant Function”. Via experiments on the recognition of English stop consonants with the method presented in this paper, it turns out reaching the desirable recognition effect.