Feature Extraction for Speaker Recognition: A Systematic Study

Abstract

The objective of a speaker recognition system is to identify a specific speaker from a spoken utterance by the speaker. Also, speaker recognition system must result in accurate identification of the speaker in a short duration. To fulfil this, extraction of the features from sound signal is an important task because speaker recognition systems are largely depend on speaker specific characteristics of a speech signal. The efficiency of this phase is crucial due to its effect on the performance and accuracy of the system. This paper presents a systematic study of the features contained in a speech signal and some techniques for extracting these features from speech signal. There are different feature extraction techniques which are used for speaker recognition like LPC, MFCC, GFCC, RASTA-PLP, etc. But MFCC and GFCC are the most widely used because they outperform other techniques and provides high accuracy rate of speaker recognition.