System Aspects of Direct Localization of Multiple Acoustic Sources Using Distributed Sensor Networks

When commonly used two-step localization techniques, such as TOA/TDOA (Time-Of-Arrival, Time-Difference-of-Arrival), DOA (Direction-of-Arrival) or RSSI (Received-Signal-Strength-Indication)/Fingerprint-based, are applied to localization in multiple signal scenarios, there is a so called data association problem. This problem is complex and complicated to solve. Data association problem is referred to grouping of localization parameters for a set of sensors (TOAs, TDOAs, DOAs, RSSI/Fingerprints) into subsets, each corresponding to a single source. With single-step (or direct) localization techniques, this problem does not exist. These techniques are of more recent date, and being numerically more demanding and complex compared to the two-step methods, they are used less frequently than two-step methods. However, they significantly outperform the two-step techniques in terms of localization accuracy and resolution properties. We present a comparative study of direct localization of acoustic signal and radio signal sources using distributed sensor networks. This study examines generic system models, generic signal models, direct localization methods, and theoretical Cramér-Rao bound (CRB) limits of location estimation error. We also include two case studies of applying direct localization of acoustic signal sources in indoor environments. These case studies involve detecting two moving e-Puck educational mobile mini-robots and detecting an acoustic signal source using a functional model of an acoustic camera. We also provide theoretical CRB limits of error estimation for these system models.