本文研究自动发音人识别中测试文本的选择.提出并证实了结合汉语特点的测试文本选择的重要性和应用潜力,并总结了几条简单规则;应用并改善了解决发音动态变化的时间域规正法. 本系统用12阶LPCC倒谱系数和基音周期构成混合特征矢量,采用三字三模板匹配的识别方法,在一般实验室环境下,以录音机为传输媒介,达到了0.6%的确认错误率。 This paper studies the choice of test texts in automatic pronunciation recognition, puts forward and confirms the importance and potential of test text selection based on Chinese characteristics, and summarizes a few simple rules. It also applies and improves the time domain rules The normal method.The system uses the 12th order LPCC cepstrum coefficients and pitch period to form a mixed feature vector, using a three-word three-template matching identification method, in a general laboratory environment, the tape recorder as the transmission medium, to achieve a 0.6% confirmation error rate .