Robust Blind Neural Network Beamformer For Cyclostationary Signals

Most communication signals have the cyclostationary properties. Many blind beamforming algorithms use cyclostationary to estimate the steering vector and adaptively obtain the LCMV optimum solution. Nevertheless, LCMV methods are sensitive to the mismatch caused by the uncalibration array or estimate error. After discussing this mismatch, a robust blind beamforming algorithm is presented in this paper. Neural network has been used widely in multidimensional signal processing, e.g. array signal processing. Implemented as a neural network, this algorithm reduces computational complexity for real time use.