| The research on sensor array processing continues to grow as new applications such as sensor networks arise. In this dissertation, we apply the array signal processing methodology to perform wideband source DOA estimation and source localization in a distributed sensor network. Theoretical analysis based on Mean Square Error (MSE) lower bounds are derived. The resulting Cramer Rao bounds (CRB) provide physical insights of the problem, which help us design effective algorithms in practice. Two types of algorithms are developed for wideband source DOA estimation and source localization: one is time-delay based and the other is based on Maximum Likelihood (ML) criterion. Efficient numerical implementations of the ML algorithm are derived and applied into a real-time sensor network testbed. Extensive simulations as well as experimental results are presented to verify the effectiveness of the algorithms.; In addition, we study various issues that we may encounter in practice, for example, source localization in a reverberant room environment. In this study, we investigate the robust algorithm design for coherent source DOA estimation.; Then, we provide a least-squares (LS) solution for source localization in a reverberant room based on our newly proposed virtual array model. The other subject is based on applying blind signal processing techniques to wideband DOA estimation. By utilizing the source statistical property, we have shown that the newly developed algorithm can outperform the ML parameter estimator. |
