基于OMAP5912的嵌入式非特定人连续语音识别系统  

Embedded Speaker-independent Continuous Speech Recognition System Based on OMAP5912

作　　者：崔金芳[1] 张雪英[1] 白静[1] 

机构地区：[1]太原理工大学信息工程学院,山西太原030024

出　　处：《电声技术》2009年第9期70-72,76,共4页Audio Engineering

基　　金：山西省科技攻关项目(2007031132);山西省留学人员回国科研项目(2006-28);太原市科技项目明星专项(08121004)

摘　　要：旨在嵌入式系统上实时实现非特定人、中等词汇量、连续语音识别。以TI OMAP5912双核处理器作为硬件平台,DSP gateway为OMAP处理器的双核通信架构,采用并行处理技术,将语音识别算法中计算密集型的任务分配给DSP处理,通过ARM和DSP协作实现语音识别,满足嵌入式语音识别系统实时性的要求。The target of this paper is to implement speaker-independent, medium vocabulary, continuous speech recognition based on embedded system timely. The hardware platform is built with TI OMAP5912 dual-core processor. For dual-core communication mechanism of OMAP processor with DSP gateway, parallel processing technology is adopted. The densely computational part is distributed to DSP, and the speech recognition algorithm is realized by ARM and DSP. Real-time request of embedded speech recognition system is satisfied.

关 键 词：嵌入式系统 语音识别 OMAP5912 DSP GATEWAY 

分 类 号：TN912[电子电信—通信与信息系统]