Research On Underdetermined Blind Source Separation Algorithm Based On Dual Microphones

Blind source separation problem originated from the “cocktail party” effect,which is characterized in that only the waveform of the source signal is recovered from the observed signal when the source signal distribution of the aliasing channel is unknown.This technology has been widely used in communication,speech,radar biomedicine and other fields.Due to limited observation conditions in practice,the number of source signals may exceed the number of observed signals.This is called underdetermined blind source separation.Aiming at the limitations of current underdetermined blind source separation,this research is based on the dual microphone underdetermined blind source separation technology.This paper uses improved independent component analysis technology?binary masking and direction of arrival estimator technology.The algorithm in this paper can better separate the speech signals.The specific work and innovation of this article are reflected in the following three points:Firstly,this paper proposes a joint similarity diagonalization algorithm based on a feature matrix with both second-order cumulants and fourth-order cumulants.The algorithm equalizes the objective function maximization problem to a joint diagonalization problem of a feature matrix that are simultaneously satisfied by a second-order delay correlation matrix and a fourth-order cumulant matrix.Compared with the joint approximative diagonalization of eigen-matrics algorithm,the improved algorithm greatly improves the separation accuracy of the algorithm.Secondly,this paper analyzes the advantages of beamforming,independent component analysis,and computer auditory scene analysis,and proposes a speech separation method based on an improved joint diagonalization algorithm combined with time-frequency masking.The improved joint diagonalization improves the separation performance,and when the source signals overlap in the time-frequency domain,the time-frequency mask can obtain a better separation effect while the calculation is simple.Finally,this paper proposes an improved two microphone underdetermined blind source separation algorithm combined with direction of arrival estimation.The improved algorithm needs to merge speech,which considers combining the direction of arrival estimation technology and time-frequency masking technology.Simulation results show that the improved algorithm can better separate source signal and reduce speech distortion to maintain speech intelligibility compared to traditional algorithms.The algorithm proposed in this paper is compared with many classical blind source separation algorithms.The evaluation results verify the superiority of the improved algorithm proposed in this paper.On the one hand,noise interference is removed,and on the other hand,the intelligibility of the speech is improved and speech distortion is reduced.