A Review On Speech Authentication And Speaker Verification Methods

Abstract

This paper gives a brief review of the techniques available for speech authentication based on edit detection and speaker verification related to application in forensics. Many past research includes the traditional methods like ENF, DFT, STFT based analysis for speech authentication while some advanced emerging trends like SVM, Ensemble learning and Convolution Neural network have been employed for speech authentication in different application. Similarly, for speaker verification, features such as LPC, MFCC, i-vector, x-vector and techniques like Neural Network, Deep learning etc have been researched. This paper summaries the various experimental results and their merits and demerits from the research papers surveyed. This review is expected to guide towards making an unbiased decision in choosing most promising method for authentication of Speaker and Speech data in forensic application.