实时语音识别系统在家庭监护机器人的实现  

A real-time speech recognition system on home care robot

作　　者：郭志强[1] 徐丽娜[1] 

机构地区：[1]吉林大学珠海学院,广东珠海519000

出　　处：《电子设计工程》2012年第7期132-134,138,共4页Electronic Design Engineering

基　　金：国家自然科学基金重点资助项目(60535010);吉林大学珠海学院资助项目(4136002556)

摘　　要：文中阐述的是家庭监护机器人项目中语音识别系统设计的部分,通过DSP、DMA和ARM Cortex-A8的并行处理,利用双缓冲的方法,在嵌入式Linux上实现了基于ATK的实时语音识别系统。文中对该系统的软硬件进行了设计。在硬件方面,给出语音识别系统的硬件组成原理,并提供了关键部分原理图;在软件方面,提出实时语音识别的方法,给出应用程序实现流程。最后通过真人说话来进行语音识别实验,实时语音识别率达到了94.67%以上,实验验证了系统的软件硬件设计的正确性。The content of this paper describes a part of the home care robot project,which is about speech recognition system.By using the DSP,DMA and ARM Cortex-A8 parallel processing,also using double buffering method and embedded Linux to realize the real-time speech recognition system based on ATK.In the hardware aspect,put forward the speech recognition system hardware composition principle,and provides a key part of schematic.On the software,a real-time speech recognition method and flow char of program on application are given.Finally,do the speech recognition experiments by some person spoken,and recognition rate is more than 94.67%,so the experimental verification of the correctness of the system design on software and hardware.

关 键 词：语音处理 家庭机器人 ATK 实时系统 

分 类 号：TP242[自动化与计算机技术—检测技术与自动化装置]