一种基于DTW的孤立词语音识别算法

针对动态时间规整(DTW)对孤立词端点检测准确性过度依赖的问题,针对上述问题,采用放宽端点和限定动态规整计算范围结合的算法,不仅更准确的放松前后端点降低端点检测的敏感度,而且结合对动态规整计算范围的限定,减少计算量,提高执行效率。分别测试了基于传统DTW算法的识别率和改进后DTW算法的识别率。实验结果表明,改进后的算法,能有效提高孤立词识别率。

A Method of Isolated Word Recognition Based on DTW

The result of isolated word recognition of the DTW depends much on the accuracy of endpoint detection.In order to reduce the influence,a modified method of endpoint detection,for relaxing the endpoint of speech signals combining with the dynamic rule computing range restriction is proposed to overcome this disadvantage.It can not only more accurately relax the forward and back endpoint to reduce its sensitivity,but also reduce the amount of calculation and improve the performance with the dynamic rule compu...