Adaptive decision-feedback generalised sidelobe canceller against desired signal and interference non-stationarity

Adaptive generalised sidelobe canceller (GSC) implemented with the least-mean-square (LMS) algorithm is a well-proven approach to effectively suppress unwanted interference. However, when the directions-of-arrival (DOAs) of the desired signal and interference sources change over time, its performance may be largely degraded because of the fact that not only the interference-cancelling filter of GSC does not properly cancel the `moving` interference, but also both the signal-matched filter and blocking matrix of GSC deviate from their best states for the `moving` desired signal. Recently, a robust adaptive decision-feedback (DF) GSC has been proposed to partly solve the problem, but if the DOAs of all sources vary significantly and continuously, the adaptive DFGSC will eventually lose track of the source motion. In this study, the authors present a new modification for the DFGSC structure, which provides additional robustness in this kind of non-stationary signal environment. Specifically, they extend the use of Householder transformation together with the LMS algorithm for DFGSC, and a new update method is introduced for the whole beamforming processor. In addition, the tracking behaviour of this specially designed adaptive structure is studied and analysed, leading to the derivation of optimal step size for the interference-cancelling filter. Simulation results show that this modified adaptive DFGSC is very effective in mitigating the described non-stationary scenario.