Performance Analysis of Linear Predictive Super-Resolution Processing for Antenna Arrays

Many applications in array processing require a high resolution beamforming method. For example, spatial division multiple access (SDMA) has been proposed to achieve higher system capacity in wireless systems. Implicit in SDMA is the ability to adaptively generate narrow antenna beam patterns, which ensures the physical separation between users and minimizes multiple user interference. However, antenna theory requires large antenna apertures to generate narrow beam patterns. This paper discusses an adaptive algorithm for extrapolating measurements from a small antenna array to a much larger virtual array. The technique utilizes linear prediction (LP) methods to perform the extrapolation. A least mean square (LMS) based algorithm was used to estimate the LP coefficients. Since the algorithm is linear, it serves both as a direction-of-arrival (DoA) estimator and a matched filter for retrieving the transmitted signal. Other popular high-resolution algorithms such as MUSIC are inherently nonlinear and thus cannot be used on their own for retrieving signal information. The performance of this algorithm is studied with respect to interference suppression, noise and spatial resolution. It is shown that the HR filter representing the LP process has poles corresponding to the DoA of the signal. The resulting LP coefficients can eliminate the interference between signals from different transmitters. However, no improvement in noise performance is seen because the receiver noise couples into the extrapolation process when two transmitters are closely located. The extrapolation algorithm is compared with the LCMV algorithm and found to produce similar results