中文摘要：

声纹识别技术作为一种新型的生物特征认证技术，英特网的快速发展给声纹识别带来很多商业上的应用，对于声纹识别技术的研究越来越受到科学和市场的重视．优化声纹识别算法速度和正确率的重要做法是提高语音信号特征参数的鲁棒稳健性．因此，本文以实验室录制的语音作为信息库，利用Mel频率倒谱系数、差分以及加权倒谱系数三个信息进行融合，采用增减阶数法分析获取高重要度的倒谱分量，组成新的融合参数，建立基于VQ的声纹识别模型，采用LBG算法设计不同码本容量的实验．通过仿真验证，本实验的倒谱分量重要度分析得到的融合特征参数在声纹识别上正确率得到提升．

英文摘要：

Voiceprint recognition technology is a new kind of biological certification technology. With the rapla development of lnternet research bring a lot of business applications for voiceprint recognition, the technology has get more and more attention of science and market. Extracting more robust speech feature parameters is the major aspect to develop the voiceprint recognition algorithm speed and correct rate. With the recording voice library in the laboratory, the paper extracted Mel frequency cepstrum coefficient and its difference, weighted cepstrum coefficient. Then get new hybrid parameters composed of parameter vector with high Importance based on analysis of increase and decrease component. Finally, the paper build a speaker recognition model based on the VQ and designed different codebook capacity experiments with the LBG algorithm. Experimental results show that hybrid parameters through this paper＇s analysis got a higher recognition rate in the speaker recognition.