给出了一种由说话者说出控制命令,机器人进行识别理解,并执行相应动作的实现技术.在此,提出了一种高准确率端点检测算法、高精度定点DSP动态指数定标算法,以解决定点DSP实现连续隐马尔科夫模型CHMM识别算法时所涉及的大量浮点小数运算问题,提高了定点DSP实现的实时性、精度,及其识别率.%In the 21st century, intelligent robot will be more and more used in every-day life. In this paper, a robot control method is presented via voice spoken by a speaker naturally. When speakers spoke commands, robot would recognize them and implement corresponding action. A new advanced voice section detection method and dynamic exponential scalar algorithm which is used to realize CHMM on fix-point DSP is given in this paper. These two methods improved robot's reality,and advanced its recognition accuracy.
展开▼