Sequential optimization: robust subspace fitting for multiple source location

An efficient algorithm designed to solve the multiple source location problem using sensor arrays is described. The algorithm is one of the class of subspace fitting techniques for source location and, in practice, it matches the performance of maximum likelihood estimation procedures. The method minimizes the error between a model of the array steering vector and the weighted eigenvectors from the measured covariance matrix of the actual array. An extension is developed for the purpose of improving estimation accuracy in the presence of sensor array errors. Simulated and experimental results are presented