Abstract: Vowels in checked syllables are almost abrupt in Chinese mandarin dialects. In order to analyze abrupt vowels in CVG syllable, the approach and methods of parametric statistical models were proposed. The method of probability distribution was used to describe the parameters. The function of parametric classification was tested by binomial and multinomial logistic models, and support vector machine was used to automatically test the classifications. The results show that the parameters of duration, DCT_1-3 are significant to the external classification, and the parameters of DCT_0-3 have significant meaning in the internal classification. The automatic testing shows that the detection rate of external classification is only 43.6% but the rate is 96.1% in internal classification. This means that abrupt feature should be additional to vowels.