Gender identification using significant Intrinsic Mode Functions and Fourier-Bessel expansion

A method to discriminate between the gender of the speakers using Intrinsic Mode Functions (IMFs) and Fourier-Bessel (FB) expansion is presented. The speech signal is decomposed into a set of Amplitude and Frequency Modulated signals called the Intrinsic Mode Functions (IMFs) using a non-linear decomposition technique called Empirical Mode Decomposition (EMD). The significant IMFs which contain most speech information are identified. They are then used for synthesizing the speech signal. This synthesized signal is segmented into frames and the FB coefficients are computed. These coefficients were used as the features for classifying the signal into male and female classes. The classification accuracy is 72.92% .