JOURNALS/BOOKS
Research on speaker gender recognition based on complex scenes 2797-2808 X Cui, C. Pan, H. Sun

Energy Education Science and Technology Part A: Energy Science and Research

2014 Volume (issues) 32(4): 2797-2808

 

Research on speaker gender

recognition based on complex scenes

 

Xuan Cui1,*, Cheng Pan2, Hua Sun3

 

1,3School of Mechanical Engineering And Automation, Xi Hua University

610039, Cheng Du ,China 02887720507,18980647911 e-mail: Cuixuan911@163.com

2 School of Electronic And Computer Engineering, Peking University ShenZhen Graduate School

 

 

 

Received: 04 March 2014; accepted: 08 May 2014

Abstract

 

This paper present a novel speaker gender recognition method under complex scene which effectively combines the method based on Mel frequency cepstrum coefficient template match and the pitch discrimination method,called The fusion method base on the contribution of MFCC and pitch. Then apply it to speaker gender recognition for noisy speech and emotional speech. The experiments indicate that the performance of this method has a great improvement compare to the method of single MFCC or pitch. In the noise environment by spectral subtraction can effectively remove the noise, frequency of male in the emotional environment changes greatly influenced by sentiment.

 

Keywords: Speaker gender recognition; MFCC; Template match; Pitch frequency discrimination; Fusion