Speaker identification system using empirical mode decomposition and an artificial neural network

This paper presents a speaker identification system using empirical mode decomposition (EMD) feature extraction method and artificial neural network in speaker identification. The EMD is an adaptive multi-resolution decomposition technique that appears to be suitable for non-linear, non-stationary data analysis. The EMD sifts the complex signal of time series without losing its original properties and then obtains some useful intrinsic mode function (IMF) components. Calculating the energy of each component can reduce the computation dimensions and enhance the performance of classification. The features were used as inputs to neural network classifiers for speaker identification. In the speaker identification, the back-propagation neural network (BPNN) and generalized regression neural network (GRNN) were applied to verify the performances and the training time in the proposed system. The experimental results indicated the GRNN can achieve better recognition rate performance with feature extraction using the EMD method than BPNN.