Improved Deep Speaker Feature Learning for Text-Dependent Speaker Recognition

摘要

A deep learning approach has been proposed recently to derive speakeridentifies (d-vector) by a deep neural network (DNN). This approach has beenapplied to text-dependent speaker recognition tasks and shows reasonableperformance gains when combined with the conventional i-vector approach.Although promising, the existing d-vector implementation still can not competewith the i-vector baseline. This paper presents two improvements for the deeplearning approach: a phonedependent DNN structure to normalize phone variation,and a new scoring approach based on dynamic time warping (DTW). Experiments ona text-dependent speaker recognition task demonstrated that the proposedmethods can provide considerable performance improvement over the existingd-vector implementation.