Feature Set Extracted Using Frequency-Time Analysis Approach for Speaker Verification

This paper compares the feature sets extracted using time-frequency analysis approach and frequency-time analysis approach for text-independent speaker verification. Mel-frequency cepstral coefficient (MFCC) feature set is extracted using time-frequency analysis approach. Temporal energy subband cepstral coefficient (TESBCC) feature set is extracted using frequency time analysis approach. The verification system is built around the likelihood ratio test, using effective GMM for likelihood functions, a universal background model (UBM) for alternative speaker representation, and using a Bayesian adaptation to derive speaker models from UBM. Results reveal that the feature set extracted using frequency-time analysis approach performs significantly better compared to the feature set extracted using time-frequency analysis approach. The equal error rates of MFCC and TESBCC feature sets are 7.19% and 3.38% respectively.