论文部分内容阅读
实现了一个仅用鼻音声母且与文本无关的汉语讲话者识别系统，根据讲话者在讲话时鼻腔相对固定、发鼻音时咽腔稳定，以及汉语鼻音声母（只有m-和n-两种）少（全部音节分别只有53和48个）的特点，使用极零（ARMA）模型获得所有汉语鼻声母音节的极点和零点系数的谱参数。系统在对20个讲话者识别时，其性能为：各个人所有单个声母测试时，总正识率为87．92％；分别随机地选用各人的人3、4、5个声母平均后测试时，则平均正识率可达91．67％、95．00％、96．67％、99．97％。
A Chinese speaker recognition system with nasal consonant and text independent is implemented. According to the relatively fixed nasal cavity when the speaker is speaking, the stable pharyngeal cavity when the nasal voice is delivered, and the initial Chinese nasal consonants (only m- and n-) (Only 53 and 48 syllables, respectively), the pole parameters and zero coefficients of all Chinese nasal vowel syllables were obtained using the ARMA model. When identifying 20 speakers, the performance of the system is as follows: the total positive recognition rate is 87.92% when all individuals are tested by one single initial consonant; the individuals who are randomly selected are 3,4, 5 consonant average post-test , Then the average positive rate of up to 91.67%, 95.00%, 96.67%, 99.97%.