Automatic Speaker Recognition using G729 Resyntesized Speech Over IP

D Yessad¹, A Amrouche²

Speech Communication and Signal Processing Laboratory
Faculty of Electronics and Computer Sciences, USTHB.

Abstract

This paper presents speaker recognition based G729 compressed speech over IP networks, the ITU-T G729 codec is investigated to encode and decode the input speech. Two methods of speech parameterization namely Linear Predictive Coding Cepstral Coefficients (LPCC) and Linear Frequency Cepstral Coefficients (LFCC) have been used as feature extractor. Speaker recognition system was designed to use those features obtained from two kinds of database: clean and G729 resynthesized speech. Experiments were performed using TIMIT database, and the effect of the G729 on GMM-UBM speaker recognition system is studied. Finally results for GMM-UBM model based both features and databases are compared and explained. It has been observed that at clean database, the system given a recognition accuracy at 89% and 87% using LPCC and LFCC successively. However the performance of both features degrades more rapidly to 80% and 77% under G729 resynthesized speech.

Keywords

• Universal Background Model (GMM-UBM), LFCC, LPCC, G729, VoIP, Resynthesized speech

Author's Profile

Author profile can be generated and linked through our partners World Book of Researchers. To include your profile online Click Here. After it is approved, please email to edlib @ asdf.res.in to create a link with all the papers.

• D Yessad : Profile
• A Amrouche : Profile

e-AID

Cite this Article as Follows

Yessad D, Amrouche A. Automatic Speaker Recognition using G729 Resyntesized Speech Over IP. International Conference on Nanoelectronics, Communications and Renewable Energy 2013. Vol. 1. Chennai: Association of Scientists, Developers and Faculties, 2013. 160-162. Print.