Dual Voice Signal Blind Separation Based On DSP

Blind source separation according to the hybrid vector data of several independent source signals obtained by observations to restore the original source signals.Normally,vector data obtained by observations is the output of a pair of sensors and the vector data received by each sensor is the different combinations of source signals,called mix signals.To blind source separation,the blind contains two different meanings.Firstly,the source signals can't be observed.Secondly,the mixed mode of source signals is unknown.Blind source separation owns great potential in many fields,such as sonar technology,radar,voice,digital image processing,bio-medical science,wireless communication,seismic observation and so on.In the field of blind source separation,the types of source signals are various.In this article,we only study the processing of voice signal.At this point,a variety of algorithms are used in blind source separation.Fast-ICA algorithm is an efficient algorithm for blind source separation,it gradually developed in recent decades,its design principle is to estimate the source signals from the mixed signals which received by sensors and driving the output signals to be mutually independent as far as possible,so that to complete the separation of mixed signals.In this article,to begin with,the Fast-ICA algorithm is determined as the separation algorithm of the system.What's more,learning the Fast-ICA algorithm and analyze it in theory.Last but not least,the Fast-ICA algorithm is simulated by MATLAB software and the results of separation is favorable.The results of separation prove the separation algorithm selected is feasible.When the separation algorithm proved to be effective,the mission is studying the hardware about the system and selecting the SEED-DEC6437 development board as the main hardware of the system.The DSP chip is DM6437 which produced by TI company and the audio processing chip is TLV320AIC23 B.Base on the original intention which is processing dual voice signals,so the system demands two sensors to receive the voice signals and two speakers to broadcast the signals of separation.The system employ the flow module of DSP/BIOS to control the data output or input.Under the cooperative processing of DSP and AIC23 B,the operation of separation has been completed.Through analysis of the results,it's clear that the mixed voice signals are separated and the results achieve the desired objectives.