The numerical simulation of the performance of a robustised, broadband, frequency domain LMS adaptive beamformer

The paper is concerned with the optimal processing of data from an array of sensors/antennas. Such sensors may be sonar, radar, VHF/HF radio or `acoustics in air´. The processing aims might be any of the following: (a) detection and identification of weak wanted signals. (b) Bearing estimation of weak wanted signals. (c) Presentation of the time series of wanted signals at maximum S/N ratio for further processing/display. (d) Accurate bearing estimation and discrimination of strong signal sources. All array beamformer techniques suffer difficulties when applied in the field. Highly optimised algorithms are quickly degraded by multipathing, array deformation and by sensor errors. Time varying noise fields and finite integration time exact a further toll. Eigenvector methods such as MUSIC are rather expensive when applied to broadband environments. The authors consider a broad band robustised LMS frequency domain adaptive algorithm as described in Nunn (1989). Its performance is analysed in numerical simulations incorporating multipathing, array distortion, sensor errors and finite integration time