摘要
用DP(动态编程)匹配算法实现语音识别,能比以前的识别简单,因而能得到快速响应。此外,由于所提取的特征向量动态地匹配给存贮器中特征模板,因而能减少由同一个字的时间多样性所造成的对精度的影响。再者,对每个模板长度限制为小于80个字节,这样,占存贮空间较省,且便于存盘。最后,介绍用语言输入来控制一个小型机器人,得到满意的结果,精度达90%。
In this paper, we introduce a DP(Dynamic Programming) matching algorithm for speech recognition. In turns out that by means of this algorithm, the procedure of recognition becomes comparatively simple so that a rapid response time can be obtained. Besides, because the characteristic vector picked out from input speech is dynamically matched with that of registered patterns, the effect on precision caused by the variety of the speech time for the same word can bc reduced. Furthermore, since every pattern is compressed into a length of less than 80 bytes, the memory locations are saved. It is also convenient to preserve the registered patterns into disks for later use. Finally the speech recognition board is connected to a host computer and a mini-robot is practically controlled by speeech input only. Satisfactory results are obtained. The experiment shows that the precision is up to 90%.
出处
《数据采集与处理》
CSCD
1989年第1期12-16,共5页
Journal of Data Acquisition and Processing
