Language and Text-Independent Speaker Recognition System Using Energy Spectrum and MFCCs

Abstract

Speaker identification, especially in critical environments, has always been a subject of great interest. In this paper, we present a language and text independent speaker identification algorithm that able to automatically identify a speaker in an audio signal with noise or real environment sound in background. The method is inspired by using a pairing of Energy spectrum and MFCCs audio feature techniques generated from base on Discrete Fourier transform (DFT). After that the audio feature extracted in real time was compared with a Euclidean Distance to measures of different between speakers to obtain the most likely speakers. The Energy spectrum feature is adopted to supplement the MFCC features to yield higher recognition accuracy for speaker identification sound.