Broadband DOA Estimation and Blind Source Separation Based on EM Algorithm

Kun-lai XIONG,Zhang-meng LIU,Zheng LIU,Wen-li JIANG,Hua-xing WANG
DOI: https://doi.org/10.3969/j.issn.0372-2112.2015.10.022
2015-01-01
Abstract:Based on the expectation-maximization (EM)algorithm,a new method for broadband direction of arrival (DOA) estimation and blind source separation is presented in this paper.First,we establish a likelihood function in the frequency domain by jointly using all frequencies information.Then,based on this function,the E-step and M-step in the frame of broadband EM algo-rithm are derived.Finally,we jointly estimate DOAs and sources through EM iteration compute.Moreover,through analyses the con-vergence of EM algorithm,we set adapted angle search space for DOA estimation,which reduce the computations of our algorithm. Compare to the traditional method,more effective information are applied in our algorithm.Therefore,the proposed algorithm has more advantages for DOA estimation and source separation.Simulation results demonstrate our algorithm is effective.
What problem does this paper attempt to address?