Source localization using wireless sensor networks

Abstract

Wireless sensors can be worn on soldiers or installed on vehicles to form distributed sensor networks to locate the source of sniper fire. A two-step source localization process is proposed for this sniper detection task. The time difference of arrival (TDOA) for the acoustic signals received by the sensors is first estimated using the generalized cross correlation (GCC) method. The estimated TDOA values are then used by the hybrid spherical interpolation/maximum likelihood (SI/ML) estimation method to estimate the shooter location. A simulation model has been developed in MATLAB to study the performance of the hybrid SI/ML estimation method. A wireless sensor network is simulated in NS-2 to study the network throughput, delay and jitter. Simulation results indicate that the estimation accuracy can be increased by increasing the number of sensors or the inter-sensor spacing. The constraint of small inter-sensor spacing on wearable sensors is found to degrade the estimation accuracy, but vehicular configuration providing larger inter-sensor spacing can help improve the estimation accuracy. The sensor topology should be well represented in all three dimensions to obtain desired estimation accuracy. The estimation accuracy is not adversely affected by sensor node failures or location perturbations. The NS-2 simulation results indicate that the wireless sensor network has low delay and can support fast information exchange needed in counter-sniper applications.