Automated pitch-based gender recognition using an adaptive neuro-fuzzy inference system

Abstract

Results on classifying a speaker on the basis of gender by processing speech and analyzing the voice samples are presented. Firstly, the speech samples are classified into voiced/unvoiced/silence by using a speech classification algorithm implemented in MATLab. The pitch of the subject's voice is extracted from the classified speech sample. Following this, automated clustering is done by an Adaptive Neuro-Fuzzy Inference System (ANFIS) to separate male and female pitch values. An automated gender classification is successfully performed by ANFIS, although, the ANFIS has to be trained before the actual classification.