Robust Beamforming by an Improved Neural Minor Component Analysis Algorithm

Recently, neural minor component analysis (MCA) has attracted much attention. It has features of high calculation speed and strong fault tolerant property. In this paper, neural MCA network is applied in adaptive beam forming. In order to get over the disadvantages that MCA based beam former cannot form null steering beams at the direction of interference and avoid stagnation behavior of algorithm convergence, a novel algorithm combining the least mean square (LMS) algorithm with MCA algorithm is provided, which uses LMS learning rule to continue the training of the neural MCA network. Simulations show that the MCA-LMS based beam former has strong stability, resembled convergence rates and can obtain a 0dB main-lobe at the direction of primary signal while forming a more than 40dB null steering at the direction of interference.