TRINICON-based Blind System Identification with Application to Multiple-Source Localization and Separation

Abstract

This contribution treats blind system identification approaches and how they can be used to localize multiple sources in environments where multipath propagation cannot be neglected, e.g., acoustic sources in reverberant environments. Based on TRINICON, a general framework for broadband adaptive MIMO signal processing, we first derive a versatile blind MIMO system identification method. For this purpose, the basics of TRINICON will be reviewed to the extent needed for this application, and some new algorithmic aspects will be emphasized. The generic approach then allows us to study various illustrative relations to other algorithms and applications. In particular, it is shown that the optimization criteria used for blind system identification allow a generalization of the well-known Adaptive Eigen- value Decomposition (AED) algorithm for source localization: Instead of one source as with AED, several sources can be localized simultaneously. Performance evalu- ation in realistic scenarios will show that this method compares favourably with other state-of-the-art methods for source localization.