孤立词语音识别系统的DSP实现  

Isolated word speech recognition system and its realization on DSP

作　　者：郑德忠[1] 宋婧婧[1] 周朝霞[1] 

机构地区：[1]燕山大学测试计量技术及仪器河北省重点实验室,河北秦皇岛066004

出　　处：《微计算机信息》2010年第26期118-119,133,共3页Control & Automation

基　　金：国家自然基金(60272027)

摘　　要：传统语音识别系统存在识别率低、接口复杂、功耗大等缺点。针对这一问题,可以采用TI公司生产的TMS320C6201B DSP开发非特定人孤立词语音识别系统。DSP主机接口部分与单片机89C51相连。由单片机控制LCD显示电路和USB与PC的通信。USB2.0总线标准数据传输速率高、接口方便。软件部分采用传统的双门限端点检测算法和DHMM识别算法。实验结果表明,该系统功耗低、体积小,并且具有较高的识别率。The conventional speech recognition system had the problems of the low recognition rate,the complicated interface and the high power loss and so on.In allusion to these problems, unspecified man isolated word speech recognition system could utilize TMS320C6201B DSP produced by TI corporation.DSP host computer interface developed the 89C51,which controled LCD and USB communicated with PC. USB2.0 bus standard had high data transmission speed and convenient interface.The software concerned double threshold endpoint detection algorithm and DHMM recognition arithmetic.The experimentation result demonstrates that the system bears the low power loss,the small volume and a high recognition rate.

关 键 词：数字信号处理 孤立词识别 离散隐马尔可夫模型 双门限端点检测算法 线性预测分析美尔倒谱参数 

分 类 号：TP912.34[自动化与计算机技术]