Comparison of Different Techniques for Gender Voice Recognition

Abstract

Gender recognition is the major research field in speech processing where different speech signals of male and female are recognized. In this paper, system developed for gender recognition using various speech signals. The male and female speech signals are recorded and it is stored in a recorded file. Fast Fourier Transform (FFT) algorithm is applied to extract the features. Power spectrum of the signal can be estimated from the FFT applied signal. The performance is evaluated by using different techniques such as energy based thresholding technique, pitch based thresholding technique and formant based thresholding technique. The performance of the system can be compared by recognition accuracy and elapsed time. This work shows that formant based thresholding for speech recognition gives better result when comparing with other methods in terms of both accuracy and elapsed time.