The Implementation of Speaker Recognition System with the Xilinx© FPGA Device

Abstract

This paper proposes a speaker identification system with a field-programmable gate array (FPGA) device of the Xilinx© Spartan III platform. The Fast Fourier Transformation (FFT) is employed to extract the features of the speeches from a group of Mandarin speakers. The Long Times Average Spectrum (LTA) is identified by calculating the Euclidean distance and matching up with the minimum distance found. Numerical experiments verify the efficiency and accuracy with the simulation in order to code the hardware digital circuits based on the proposed model. This speaker identification system applying the FPGA for a feature extraction and pattern recognition has been implemented. Test signals were generated using ModelSim© and Matlab© software. Comparison with results in Matlab© simulation, the proposed system can achieve a pretty good accuracy on identifying speakers with the FPGA device.