| 摘要: |
| 对基于隐马尔可夫模型(HMM)的说话人辨认系统进行了讨论，完成了系统设计。对系统中矢量量化这一关键性环节进行了改进，提出了一种新的基于遗传算法的码本生成方法。测试结果表明，改进后的系统具有较高的正确识别率，特别是在与文本无关的情况下。 |
| 关键词: 语音信号处理,说话人识别,说话人辨认,隐马尔可夫模型,矢量量化 |
| DOI：10.3969/j.issn.1001-893X. |
| Revised:May 26, 2003 |
| 基金项目:江苏省"青蓝工程"跨世纪学术带头人专项基金资助(QL003YZ)，南京邮电学院科研项目(《说话人识别的方法与技术》)资助 |
|
| HMM-based Speaker Identification System and its Improvement |
|
| () |
| Abstract: |
| The speaker identification based on Hidden Markov Model(HMM)is introduced. As an important part, the vector quantization(VQ)is modified. A new method based on genetic algorithm is proposed to design the codebook.Experiment shows the identification accuracy of the system is improved, especially for text-independent speaker recognition. |
| Key words: Speaker recognition,Speaker identification,Hidden Markov Model(HMM),Vector Quantization(VQ) |
