基于HMM语音识别技术在ARM平台的实现  

HMM-based Speech Recognition Technology in Realization of ARM

作　　者：祝常健[1] 胡维平[1] 叶佳宁[2] 

机构地区：[1]广西师范大学物理与电子工程学院 [2]广西师范大学计算机科学与信息工程学院

出　　处：《微计算机信息》2009年第35期143-145,共3页Control & Automation

基　　金：基金申请人:胡维平王修信;项目名称:基于HHT变换和连续语言声的艺术嗓音客观检测与评价研究;基金颁发部门:广西自然科学基金委(桂科自:0448035)

摘　　要：语音识别技术越来越广泛的应用于嵌入式系统平台。在分析基于HMM语音识别原理的基础上,设计并实现了一个基于ARM9和嵌入式Linux系统的特定人、孤立词和小词汇量的嵌入式语音识别系统。采用S3C2410微处理器和嵌入式Linux操作系统,将交叉编译后的语音识别C语言程序移植入嵌入式Linux操作系统的文件系统,实现语音识别系统的功能。给出了系统整体的软硬件框架,并给出了实时输入识别0到9的汉语数字语音识别的实验结果,探讨了在不同码本长度和不同隐含马尔可夫模型状态数的识别性能,为语音识别技术在嵌入式设备实际应用提供参考。Speech recognition technology increasingly extensive used in embedded system platform.Designed and implemented a embedded speech recognition system which base on ARM9 and embedded Linux system of the specific,isolated words and small vocabulary.Used S3C2410 microprocessors and Linux operating systems,Transplant the C language of speech recognition program to the embedded Linux operating system’s file system when had been cross-compiled. Gave the system s hardware and software framework,and gave the experiments results which are real-time recognition of the 0-9 Chinese figure speech,explore the recognition performance when in the different code book length and different state numbers of the Hidden Markov Model, provision the speech recognition technology for embedded devices in the practical application＇s reference.

关 键 词：语音识别 嵌入式系统 隐含马尔可夫模型 S3C2410处理器 

分 类 号：TP391.42[自动化与计算机技术—计算机应用技术]