High Impact Factor : 4.396 icon | Submit Manuscript Online icon |

Speaker Identification Based on Voice Samples using MFCC and GMM

Author(s):

Avinash Kumar , Dayananda Sagar College of Engineering; Ankit Raj, Dayananda Sagar College of Engineering; Avantika Shee, Dayananda Sagar College of Engineering; Kundan Kumar, Dayananda Sagar College of Engineering; Nagarathna R, Dayananda Sagar College of Engineering

Keywords:

Speaker Identification, Feature Extraction, Feature Identification, Mel-Frequency Cepstral Coefficients (MFCC), Gaussian Mixture Model (GMM)

Abstract

Speaker Identification is the task of claiming a speaker’s identity based on the characteristics contained in the voice signal. The identification is done by dividing the task in two phases named Feature Extraction and Feature Identification. Various methods are available for extraction of a voice signal such as Mel Frequency Cepstral Coefficients (MFCCs), Linear Frequency Cepstral Coefficients (LFCCs), Linear Predictive Cepstral Coefficients (LPCC), Perceptual Linear Prediction Cepstral Coefficients (PLPCCs) etc. Similarly, modern methods available for identification include Vector Quantization (VQ), Hidden Markov Model (HMM), Gaussian Mixture Models (GMM) etc.In this paper, based on our survey on the grounds of noise reduction, computational time, accuracy and efficiency, we propose that an MFCC-GMM model is best suited for identifying a speaker.

Other Details

Paper ID: IJSRDV3I31208
Published in: Volume : 3, Issue : 3
Publication Date: 01/06/2015
Page(s): 3326-3328

Article Preview

Download Article