Composite wavefront decomposition via multidimensional digital filtering of array data

A solution is obtained to the problem of estimating the number, vector velocity, and waveshape of overlapping planewaves in the presence of interfering planewaves and channel noise, where previous solutions have assumed one or more of these quantities as known. A general optimum solution is not found; instead, a heuristic solution is presented along with a complete working implementation program for large scale computers. For the case where the number of waves and the vector velocities are known, the solution is optimum. The detection of waves and the estimation of their bearing, velocity, and waveshape is accomplished via digital filtering of the frequencywavenumber power spectrum, which is computed via an efficient estimator, of the array sensed data. A new approach to the multiwave estimation problem is to reduce it to a succession of single wave problems using especially developed frequency-wavenumber filters. Special attention is given throughout the study to computationally efficient approaches. The results of the paper are placed in perspective by showing how the historically important approaches to the processing of array data such as delay and sum, weighted delay and sum, array prewhitening, beam forming, inverse filtering, least mean-square estimation, and maximum likelihood estimation are related via the spatio-temporal filtering of the frequency-wavenumber spectrum. The spectral estimation, digital filtering, and the multiwave maximum likelihood estimator developments are demonstrated by the processing of a set of simulated planewaves of various bearings, velocities, and frequencies, as well as by processing electroencephalographic (brain wave) data monitored via an array of scalp electrodes.