Comparative Analysis of Prosodic Features and Linear Predictive Coefficients for Speaker Recognition Using Machine Learning Technique

Author

Baidwan, V.S. ; Gujral, S.

Author_Institution

Dept. of Comput. Sci. & Eng., Chandigarh Univ., Mohali, India

fYear

2014

fDate

12-13 Sept. 2014

Firstpage

1

Lastpage

8

Abstract

Speaker recognition is a biometric identification method that uses different features of individual´s voice for automatically identifying a speaker among a population. Two different features set for text dependent speaker recognition. A comparison is performed between Linear Predictive Coefficients (LPC) and Prosodic Features (F₀, F₁, F₂, and F₃) along with Radial Basis Function Network (RBFN) for recognizing a speaker population of 100 speakers. The results conclude that prosodic features performed better than LPC in terms of accuracy, precision and recall.

Keywords

radial basis function networks; speaker recognition; F₀ prosodic feature; F₁ prosodic feature; F₂ prosodic feature; F₃ prosodic feature; LPC; RBFN; accuracy value; automatic speaker identification; biometric identification method; feature set; linear predictive coefficients; machine learning technique; precision value; radial basis function network; recall value; text dependent speaker recognition; voice features; Accuracy; Cepstrum; Feature extraction; Mel frequency cepstral coefficient; Speaker recognition; Speech; Speech recognition;

fLanguage

English

Publisher

ieee

Conference_Titel

Devices, Circuits and Communications (ICDCCom), 2014 International Conference on

Conference_Location

Ranchi

Type

conf

DOI

10.1109/ICDCCom.2014.7024705

Filename

7024705