Abstract:

This paper considers a parametric approach for adaptive multichannel signal detection, where the disturbance is modeled by a multichannel auto-regressive (AR) process. Motivated by the fact that a symmetric antenna geometry usually yields a per-symmetric structure on the covariance matrix of disturbance, a new per-symmetric AR (PAR) modeling for the disturbance is proposed and, accordingly, a per-symmetric parametric adaptive matched filter (Per-PAMF) is developed. The developed Per-PAMF, while allowing a simple implementation like the traditional PAMF, extends the PAMF by developing the maximum likelihood (ML) estimation of unknown nuisance (disturbance-related) parameters under the per-symmetric constraint. Numerical results show that the Per-PAMF provides significantly better detection performance than the conventional PAMF and other non-parametric detectors when the number of training signals is limited.