THIRD ANNUAL
ASAP '95 WORKSHOP

Maximum-Likelihood
Techniques for Joint
Demodulation and
Beamforming with
Adaptive Antenna Arrays

Abstract A maximum-likelihood (ML) formulation of waveform demodulation for narrowband signals received at multiple sensors is presented. The signals are received in unknown Gaussian cochannel interference which can have a mix of diffuse and specular components. The signal wavefronts are assumed unknown in the statistical formulation of the problem. Thus, beyond matching receiver channels to maintain null depths, no sensor calibration is required.

Demodulation occurs jointly with beamforming. For one signal in unknown interference, the solution to the ML formulation can be expressed in terms of an iterative search over candidate beamformers involving demodulation of the beamformer output. A specific quality measure comparing the raw beamformer output with the remodulated signal is used to adjust the beamformer. For several signals in unknown interference, the ML formulation reduces to an estimation/subtraction procedure that is built from the one-signal ML approach and is applied to each of the signals sequentially.

Several simulations of beamforming/demodulation, involving different types of waveforms (e.g., known, constant envelope, and quadrature amplitude modulated), are presented. In some cases it is possible to achieve performance close to that of an ideal beamformer/demodulator, which has precise knowledge of the entire signal environment excluding the data content of the signal.