A neural minor component analysis algorithm for robust beamforming

A novel minor component analysis (MCA) learning rule is presented which includes a penalty term on the self-stabilizing MCA learning rule. After a presentation of convergence and steady-state analysis, it is shown how the novel MCA learning rule can be used for realizing robust constrained beamforming. Constrained beamformer power optimization principle is employed, which allows to improve the performance of the beamforming algorithm by emphasizing white noise sensitivity control and prior knowledge about the disturbances. Computer simulations show the novel MCA learning rule has strong stability, resembled convergence rates and real-time signal tracking ability, compared with the first minor component analysis (FMCA) learning rule.