| 说话人辨认系统中码本生成的新算法 |
| |
| 引用本文： | 张玲华,石操,张子菁,杨震,郑宝玉.说话人辨认系统中码本生成的新算法[J].信号处理,2004,20(4):391-394. |
| |
| 作者姓名： | 张玲华 石操 张子菁 杨震 郑宝玉 |
| |
| 作者单位： | 南京邮电学院信息工程系,南京,210003 |
| |
| 基金项目： | 江苏省教育厅青蓝工程项目，南京邮电学院校科研和教改项目 |
| |
| 摘 要： | 提出了一种新的适用于离散HMM说话人辨认系统的VQ码本训练方法,码本的训练准则是使码本中各码字的利用率趋于均等.将新方法训练的码本与用LBG算法训练的码本进行了比较,实验表明,在基于离散HMM的说话人辨认系统中,用新方法训练的码本性能优于用LBG算法训练的码本,特别是在与文本无关的情况下,使系统的正确辨认率显著提高.
|
| 关 键 词： | 说话人识别 说话人辨认 隐马尔可夫模型(HMM) 矢量量化(VQ) 遗传算法(GA) |
| 修稿时间： | 2003年5月27日 |
A Novel Method to Train VQ Codebook for Speaker Identification |
| |
| Zhang Linghua Shi Cao Zhang Zijing Yang Zhen Zheng Baoyu.A Novel Method to Train VQ Codebook for Speaker Identification[J].Signal Processing,2004,20(4):391-394. |
| |
| Authors: | Zhang Linghua Shi Cao Zhang Zijing Yang Zhen Zheng Baoyu |
| |
| Abstract: | A novel method is proposed to trained VQ codebook for HMM-based speaker identification. The codebook is trained based on the criterion of making each codeword in the codebook to share training vectors equally. An evaluation experiment has been conducted to compare the codebook trained by the Linde-Buzo-Grey(LBG) and the new algorithm. It is showed that the codebook trained with new algorithm can give a much higher speaker identification rate than the LBG trained codebook when used in HMM-based speaker identification, especially for text-independent speaker identification. |
| |
| Keywords: | speaker recognition speaker identification hidden markov model (HMM) vector quantization (VQ ) genetic algorithm |
| 本文献已被 CNKI 万方数据 等数据库收录！ |
|
