Analysis of Classifiers for Gender Identification using Voice Signals

Abstract

This paper presents analysis of different classifiers to identify gender-based voice signals using MFCC. Gender is an important and critical parameter of individuals. The gender of the human being can be identified with the help of voice signals. To Communicate between human beings, speech is the most convenient and suitable medium. Speaker produces the voice signals with the help of a vocal cord for shouting, singing, talking etc. The vocal cord is the most essential part to produce voice signals. The analysis of the voice signals plays a very important role to identify the gender for authentication in highly secure areas. For the recognition of the speaker, the gender identification has important because it reduces the search space. In this paper, Mel Frequency Cepstral Coefficients (MFCCs) are used as a component of a feature vector to recognition of the speaker's gender. Support Vector Machine (SVM), Vector Quantization (VQ), Multilayer Perceptron (MLP), Gaussian Mixture Model (GMM) and Learning Vector Quantization (LVQ) and k-Nearest Neighbor (KNN) are used as a classifier to classify the speaker with the help of voice signals of the speaker. The highest accuracy for the identification of gender is 99.5% using the Support Vector Machine (SVM).