Text dependent speaker verification using Algebraic Approach (AA) method and DTW under limited data condition

Text dependent speaker verification using Algebraic Approach (AA) method and DTW under limited data condition Speaker verification is the process of automatically authenticating the identity of the speaker in order to protect the resources by the method of access control. Usually identity is claimed by presenting a unique personal possession, here in this case a sample of the user’s voice. This paper presents an approach of text dependent speaker verification using MFCC as feature vector with Algebraic Approach Method (AA) and DTW as modeling technique. A database of 30 Speakers was collected over practical noisy environment and was used for testing and validating the modules. The experimental results suggest that using the combined techniques of DTW and AA at the decision level, the total success rate (TSR) of the system increases to 85.445 % which provides a relative improvement of 0.675 % in contrast to individual classifiers. The approaches are described and detailed experimental results and analysis are presented and discussed.