Acoustic-based localization in Wireless Sensor Networks using MISO least squares estimators
In this thesis, a solution to the localization problem in Wireless Sensor Networks (WSN) is presented. For many applications knowledge of exact position of the sensor nodes is of primary importance. The measurement of fields (like temperature, light or pressure) is analogous to a sampling problem in 2-D or 3-D. The estimation of field values in areas where no sensor is available requires precise knowledge of the neighboring nodes in order to perform field interpolation. Detection and tracking of event moving through a WSN has no significance without the knowledge of the sensor locations. Finding or tracking people on a building, animals in the wilderness and products on a store are classical localization problems that could use WSNs. We propose a method to estimate sensor node location on a WSN using acoustic signals. The proposed method is inspired by work in the multimedia research community where a speaker needs to be located and tracked using speech signals and microphone arrays. For WSNs, this method assumes that sensor nodes are equipped with hardware to produce and to acquire acoustic signals. This assumption conforms to the current technical capabilities found in WSN systems. In addition, a protocol that combines RF communication, actuation and sensing is necessary to achieve localization across the network. The localization scheme starts with a set of unknown (i.e., not localized) sensors and a set of reference sensors with known location (determined by other means). Each unknown sensor produces a wideband acoustic signal that is received by the set of reference sensors. The difference in signal travel time between the source and reference nodes are measured to obtain the Time Difference of Arrivals (TDoAs). TDoAs are estimated using one of these methods: Cross-Correlation (CC), Phase Transform (PHAT), and Magnitude Difference (MD). Once we have the TDoAs, the second step is to estimate the location based on the TDoAs. This estimation is performed by using techniques of Least Square Estimation (LSE) such as Spherical Intersection (SX), Spherical Interpolation (SI), and Global Spherical Least Square (GSLS). The combination of these algorithms is studied in this thesis. We determine the best combination of TDoA and LS location estimators given conditions commonly found in WSNs deployed on open fields.
Hermosillo, Jesus Manuel, "Acoustic-based localization in Wireless Sensor Networks using MISO least squares estimators" (2008). ETD Collection for University of Texas, El Paso. AAI1456747.