Efficient Time of Arrival Calculation for Acoustic Source Localization Using Wireless Sensor Networks

Acoustic source localization is a very useful tool in surveillance and tracking applications. Potential exists for ubiquitous presence of acoustic source localization systems. However, due to several significant challenges they are currently limited in their applications. Wireless Sensor Networks (WSN) offer a feasible solution that can allow for large, ever present acoustic localization systems. Some fundamental challenges remain. This thesis presents some ideas for helping solve the challenging problems faced by networked acoustic localization systems. We make use of a low-power WSN designed specifically for distributed acoustic source localization. Our ideas are based on three important observations. First, sounds emanating from a source will be free of reflections at the beginning of the sound. We make use of this observation by selectively processing only the initial parts of a sound to be localized. Second, the significant features of a sound are more robust to various interference sources. We perform key feature recognition such as the locations of significant zero crossings and local peaks. Third, these features which are compressed descriptors, can also be used for distributed pattern matching. For this we perform basic pattern analysis by comparing sampled signals from various nodes in order to determine better Time Of Arrivals (TOA). Our implementation tests these ideas in a predictable test environment. A complete system for general sounds is left for future wor