摘要
介绍了一款非特定人嵌入式语音控制系统。硬件平台采用mini2440开发板,其CPU处理器是三星公司的S3C2440;针对嵌入式系统存储空间有限和实时性要求高的特点,在特征提取后加入LBG算法以减少系统的存储量和计算量;识别阶段采用DHMM算法,计算量少,适应性强。将该系统用于语音对机械手臂的控制试验,结果表明本系统的识别率达到了98%以上,运算速度快,计算复杂度小,能满足一般应用对语音控制的需求,在市场上具有广泛的应用前景。
A speaker independent embedded speech system is introduced. The hardware platform is mini2440 development board using S3C2440 microprocessors; the storage space is limited and real-time is demanded highly in embedded system. In order to resolve the problems, LBG algorithm is applied after feature extraction to reduce storage of system and calculation, and HMM algorithm is used in recognition, less computational effort and strong ada- ptability. The speech recognition accuracy reaches up 98% when using the system to do test to control the mechanical arm. The system can meet the accuracy and time requirements of speech recognition system, It has great potential in market application.
出处
《电声技术》
2010年第10期54-56,61,共4页
Audio Engineering
