基于仿生模式识别的非特定人连续语音识别的研究  

Speaker-independent continuous speech recognition based on the biomimetic pattern recognition

作　　者：叶虹[1] 

机构地区：[1]浙江工业大学浙西分校,浙江衢州324000

出　　处：《浙江工业大学学报》2006年第4期433-435,共3页Journal of Zhejiang University of Technology

摘　　要：仿生模式识别基本原理应用到非特定人连续语音识别中,在构建语音模型时,将高维空间点覆盖的理论应用到非特定人的连续语音识别的学习过程中,在识别时,采用高维空间覆盖动态搜索的方法,对连续语音进行动态搜索,这样就可以避免因分割错误而带来的错误识别.通过实验得到了较为满意的识别结果.We use the theory of biornimetic pattern recognition for the speaker-independent continuous speech recognition. In the process of constructing the speech model, the high-dimension space covering theory （HDSCT） is applied to the study process of the speaker-independent continuous speech recognition. The dynamic searching theory of high dimensional space vertex covering is used to search the rolling time axis of the continuous speech dynamically in order to avoid the wrong division and wrong recogrition. We get a satisfied result through the experimentation.

关 键 词：仿生模式识别 高维空间 连续语音 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]