Variable Sliding Window DTW Speech Identification Algorithm

Dynamic Time Warping (DTW) is a kind of simple and efficient method in speech identification. The human pronounces with variable pitch period at different time. Speech parameter extraction with fixed-length window will cause error discrimination to some extent because of different signal period in windows. Sliding windows with variable length is to select sliding windows with different length according to the variation of the pitch period of the speech signal in order that each window contains speech signals with identical period.Speech segmentation via sliding windows with variable length can eliminate the errors caused by the difference of speech signals contained in each window. Experimental results verify the improvement on the precision of speech identification.