The research addresses estimation and tracking of direction of arrival @OA) and associated parameters of narrowband and wideband signals impinging on a uniform linear array of sensors. The signals are modeled as sample functions of a Gaussian stochastic process. Computationally efficient, approximate maximum likelihood (ML) methods are developed for direction of arrival estimation of narrowband signals impinging on a large array of sensors. A new likelihood function is formulated based on a large M (# sensors) Taylor's series approximation of the original likelihood function. Asymptotic expressions for Cramer-Rao lower bounds on the DOA estimates are derived. From the positive definiteness property of the Fisher information mamx, a resolution criterion for closely spaced sources is proposed. An algorithm for tracking multiple narrowband signal sources in near-field is proposed based on joint estimation of angle and range by the maximum likelihood principle. For sources modeled as wideband signals, a new scheme for tracking direction of arrival is proposed. The wideband signals are modeled as vector auto regressive models so that their spectral densities are characterized by a finite number of parameters. A Bayes classifier is employed for data association. A new method is proposed for tracking and data association by estimation of singularity of higher order curves fitted to data @OA estimates). At every tracking time instant, the intercept point forecast information of pairs of signal tracks obtained from existing track data is employed for data association. The forecasted intercept point is recognized as the estimated singularity of a single second order curve fitted to data from every pair. Data association is achieved by detecting cross-over from the knowledge of these forecasts, and by suitable evidence combination of cross-over detection.

Date of this Version