Linear Discriminant Analysis F-Ratio for Optimization of TESPAR & MFCC Features for Speaker Recongnition

This paper deals with implementing an efficient optimization technique for designing an Automatic Speaker Recognition (ASR) System, which uses average F-ratio score of TESPAR(Time Encoded Signal Processing And Recognition) and MFCC(Mel frequency Cepstral Coefficients) features, to yield high recognition accuracy even in adverse noisy conditions. A new ranking scheme is also proposed in order to stabilize the rank of features in various noise levels by taking Arithmetic Mean of the F-Ratio scores obtained from various levels of Signal to Noise Ratio (SNR). The result is presented for a Text-Dependent ASR system with 20 speaker database. An RBF (Radial Basis Function) Neural Network is used for Recognition purpose. Also a comparative study has been performed for recognition accuracies of optimized MFCC and TESPAR features and we conclude that new proposed average F-Ratio technique has resulted in better accuracy compared to simple F-ratio in noisy environment and also we came to know that TESPAR features are more redundant compared to MFCC. Index Terms ASR, F-Ratio, Average F-Ratio, TESPAR, RBF Neural Network, MFCC.