Comparison of DTW score and warping path for text dependent speaker verification system

Abstract

Dynamic Time Warping has always been a popular technique for pattern matching of two speech samples for Automatic Speaker Recognition. DTW score evaluated from the minimum distance matrix is generally used to identify the similarity between two speech samples and thereby giving a decision for the Text Dependent Speaker Verification system. This paper discusses a system based on DTW trace back path for the alignment of two speech samples for the decisions are thereby compared with the performance of the system with the system based on DTW Score. The result shows a significant improvement in the performance of the system suggesting that the trace back path provides a better similarity measure as compared to that of the DTW Score.