Comparative Analysis of Gender Identification using Speech Analysis and Higher Order Statistics

Gender identification via speech processing is one of the hot research topics among the security research community. Many cyber systems are being developed to recognize human speech type. These systems mainly comprise of a feature segment process which extracts and selects the features of human speeches. Feature extraction and feature selection are the most noteworthy phase of speech recognition involving numerous strategies. The purpose of this paper is to investigate the potential effectiveness of spectral analysis and higher-order statistics performed over the speech segments of different genders. Spectral analysis is done via spectral descriptors consisting of varied parameters which are widely used in machine learning applications. The varied gender speeches are distinguished by means of parameters, i.e., higher order statistics, like spectral centroid, spectral entropy, spectral kurtosis, spectral slope and spectral flatness. The results obtained show successful discrimination of male and female speeches based on the peakiness of speech, voiced and unvoiced and higher and lower formants.