实现了一个嵌入式平台上的轻量级语音识别程序库
) 摘要: 本文针对当前嵌入式平台的浮点计算能力不高与资源有限的情况
对现有的语音识别通用算法进行了一些改进
) Abstract: To the problem of the poor quality of float point computability and poverty of resource of the current embedded platforms, this paper proposed a suite of improvement to the generic speech recognition algorithms, and implements a light weight level speech recognition library for embedded platforms
By converting the float point arithmetic to the fixed point arithmetic, plus compressing of the resource requirement, this paper archives usable robust and real-time performance without obvious correctness loss
The proposed library has been simulated on the Matlab platform and implemented on the Windows Mobile platform [HP rx1950]
Mostly dependencies library are generic and portable, so porting to another embedded platform would require little extra effort
Keywords: ASR;embedded platform;fixed point arithmetic;windows mobile 下载PDF阅读器 PDF全文下载: 初稿 ( 224 ) 作者简介: 通信联系人: 【收录情况】 中国科技论文在线: 湛宗儒
一个嵌入式平台语音识别程序库的实现[EB/OL]
北京:中国科技论文在线
嵌入式平台
通过把浮点运算转化为定点计算
移植到其他嵌入式平台也是切实可行的
程序库在Matlab中仿真通过并且在Windows Mobile平台[HP rx1950]上实现