Abstract: Describes briefly a study of a few fundamental problemson Speaker Recognition. Speech parameters are the base and a speech database is needful for speaker recognition. Our study used VQ technic and materiel of speech database which created by us. It is shown by experiments that in some introduced parameters a mixed parameter is the best, secondly is cepstral coefficient.

Key words: speaker recognition, speech parameter, VQ, cepstral coefficient, linear predictive coding

摘要： 本文对说话人识别中的几个基本问题进行了研究。语音参量是说话人识别的基础，用矢量量化方法，使用自建的语音库中的材料，研究了说话人识别中的各种参量的效果。实验表明，所采用的参量中，一种混合参量MC最好，倒谱系数CE次之。

关键词: 说话人识别, 语音参量, 矢量量化, 倒谱系数, 线性预测编码