| China is a country with multi-nation and many kinds of language. But as a country she has a common and standard language—Mandarin. With the development of the society and economy, it shows more importance to popularize mandarin. So we need a method to judge what tone is standard or to know how to pronounce exactly. Presently the main way is based on subjective assessment. But it has many shortages such as different assessment by different judge, too much subjective ingredient in such assessment and so on. Therefore if there is an objective assessment, there will be the same judgment. It will also save much time, because this method can implement in the computer. For all above reasons the project comes up. This paper focuses on the objective assessment of single word. In present there is no same research and available theories. Base on a large amount of speech study and experiments, this paper gives a new method of computer aided assessment of mandarin pronouncing accuracy。The first, By dividing between consonant and vowel of Chinese syllables, we establish a speech distinction system of Chinese pronouncing accuracy using the continuous Gaussian mixture HMM. The second, we establish a speech accuracy system of Chinese using the RBF network. On the choice of unit, unit structure, and unit parameters, the paper gives a thorough consideration for the properties of Chinese speech Using Mel cepstrum and Bark-SD At last, adding the tone and the length of initial-like units to finish mapping from the characteristics to the expert's score. The experimental results showed good performance... |