Dynamic time warping for speech recognition with training part to reduce the computation

Dynamic time warping (DTW) is a popular automatic speech recognition (ASR) method based on template matching[1], [2]. DTW algorithm compares the parameters of an unknown word with the parameters of one reference template. But the recognition rate is limited. To increase the number of reference templates for the same word will improve the recognition rate, but it will lead to spend a lot of computing time and memory resource. In this paper we proposed a method to reduce the number of reference templates, thus reduces the computing time and memory resource and also keep the high recognition rate.