Low-Complexity Audio Signal Processing for Localization in Indoor Scenarios

During the last years, wireless sensor networks (WSNs) have received significant attention from the research community, as one of the emerging technologies for the new millennium. A WSN is composed by many (e.g., hundreds) devices with limited processing and communication capabilities. Therefore, energy saving is one of the major issues and information processing has to be performed with low complexity. Possible applications for WSNs are surveillance, environmental monitoring, flow control, etc, and it may be possible to work in indoor scenarios [1]. The application of interest in this paper is the localization of a person or an object in indoor scenarios [2], but our approach is also suitable for outdoor scenarios. We will refer to the person (or object) to be localized as the entity. In these scenarios, it is realistic to assume that the nodes are not equipped with global positioning system (GPS) devices and, therefore, other techniques are needed to perform efficient localization. Most of the works about localization in sensor networks are based on the assumption that the entities are equipped with devices which radio-communicate with some reference nodes (denoted as anchors). The positions of the anchors are supposed to be known and the position of the entity of interest is inferred by “combining” the information available at each anchor (e.g., by triangulation). In the literature, several techniques, based on different methods, have been proposed to obtain a sufficiently low estimation error [3, 4]. The computational complexity of these algorithms is a crucial issue for WSN-based applications. In [5], the authors propose a sub-optimal hierarchical algorithm, which solves the localization problem without resorting to the optimum maximum likelihood (ML) technique, whose computational complexity becomes too high to be of any practical interest. In [6], an adaptive approach to localization problems, obtained by solving a sequence of very small optimization subproblems, is considered.