A Covariance Modeling Approach to Adaptive Beamforming and Detection
Abstract
The subject of this report is the general problem of signal processing for sensor arrays. Under certain reasonable assumptions, the model for the noise covariance matrix of the vector of array outputs is an integral involving the spatial-temporal power-spectral-density function. This report examines the application of this covariance model to problems in adaptive beamforming and detection. A constant false alarm rate detector, based on unconstrained maximum-likelihood techniques, is derived and analyzed. Techniques such as this do not fully exploit the data model and can show an appreciable loss in performance compared to optimal techniques. The space of noise covariance matrices possible from a particular array is characterized, yielding representations for the space and members of the space in terms of finite numbers of spectral points. These representations are used to derive constrained maximum-likelihood estimators that jointly estimate the parameters of the density function. Two approaches that use the constrained covariance estimates to perform beamforming are described and compared. The loss in signal-to-noise ratio and the variance of the estimators are shown to be less for these approaches than for those that do not use the covariance model. Detection methods based on the generalized likelihood ratio test and a constant false alarm rate matched-filter detector are analyzed, and simulation results are presented.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jul 30, 1991
- Accession Number
- ADA241887
Entities
People
- Frank C. Robey
Organizations
- Massachusetts Institute of Technology