基于TMS320C6678的语种识别并行算法设计与实现  

Parallel design and implementation of language recognition based on TMS320C6678

作　　者：吉立新[1] 刘伟伟[1] 李邵梅[1] 

机构地区：[1]国家数字交换系统工程技术研究中心,河南郑州450002

出　　处：《电子技术应用》2012年第10期37-40,共4页Application of Electronic Technique

基　　金：国家863计划重点项目(2011AA010603)

摘　　要：针对语种识别在实际应用中的实时性需求,在分析语种识别算法原理和多核DSP任务并行的基础上,设计了一种基于TMS320C6678的语种识别算法并行实现方法。该方法实现了任务级并行流水和核间高效通信。实验结果表明,该方法可靠、有效,并且使该系统至少能够完成132路的实时识别任务。Aimming at the real-time requirement in practical application of the language recognition, based on the analysis of language recognition and parallel task in multicore DSP, this paper designs a method of language recognition algorithm parallel im- plementation based on TMS320C6678. This method implements task-level parallel pipeline and efficient inter-core communication. Experimental results show that the method is reliable and effective, and the language recognition system based on TMS320C6678 is able to fulfill the 132-access recognition tasks at least.

关 键 词：语种识别 多核DSP TMS320C6678 并行处理 

分 类 号：TN302[电子电信—物理电子学]