A Review on Extraction of Speaker Related Variability Features Using Short-Term Feature Extraction Techniques

Abstract

In the rapidly developing digital environment, speaker verification is becoming more and more popular with prominent applications in security, automation, and authentication. Techniques for verifying speakers reject the brief fluctuations in the feature extraction stage that contain significant speaker-related characteristics. This page discusses a variety of feature extraction methods, ranging from short-term to long-term characteristics. Also the article discusses about speaker related variability features extraction with various fusion schemes such as Mel-Frequency Cepstral Coefficients (MFCC), Frequency Domain Linear Prediction (FDLP), Mean Hilbert Envelope Coefficients (MHEC) and Power-Normalized Cepstral Coefficients (PNCC) which acts as a complimentary for extracting common short-term speaker-related features. To combat accurate speaker verification, the review article with give a base idea for extracting short-term features for accurate recognition of speaker by compacting classifier efficiency.