A new gender detection algorithm considering the non-stationarity of speech signal

Abstract

This paper presents a new pitch based gender detection algorithm by analyzing the non stationary behavior of speech signal. To find pitch a peak detection algorithm is developed which find the dominant frequencies of vowel part of speech signal and then select the fundamental frequency from them. The evaluation is done on the 200 voice samples by using POC as the accuracy parameter. Gender identification is an important step in speaker and speech recognition system. A gender dependent system reduces the size and complexity of the system. Detecting the gender from the non linguistic characteristics of the voice is well known as gender detection. To extract the gender information from the speech signal we have used a feature called pitch from voiced part of the speech signal. To extract the mentioned feature ‘pitch’ from the speech signal we develop a peak detection algorithm and used standard Fast Fourier Transform (FFT) technique. Further, this feature is used for the speaker's gender classification.