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.
展开▼
