Sensor Networks for Acoustic Source Localization Using Acoustic Fingerprint in Urban Environments and Construction Sites

Abstract

We investigate the feasibility and performance of the acoustic localization system using the acoustic fingerprint for asynchronous wireless sensor network (WSN) in urban environments and construction sites. The location estimation is calculated by comparing the acoustic fingerprint obtained from multiple sensors with those pre-computed in the database. To calculate the fingerprint, we avoid expensive measurement process by using a 2-dimensional finitedifference time-domain (FDTD) to approximate acoustic propagation in urban area. The implementation cost for constructing the acoustic fingerprint map using FDTD is small compared to that of the exhaustive data measurements over an entire test site. We select the direction-of-arrival (DoA) of the first arrival path of the acoustic signal as a location fingerprint to avoid complexity from synchronization. The fingerprint from each node is weighted by the measured amplitude of the received acoustic waveform to take into account the decrease in the amplitude due to the distance between the source and the sensor node. We test our proposed localization algorithm at the 140x80 m artificial village area used for military drills. With proper node placement, our proposed algorithm can achieve strong localization performance using small number of sensor nodes. In particular, the root-mean-square-error (RMSE) between the estimated and accurate source position is 6.30 meters using observations from only 3 sensor nodes. Our proposed algorithm exhibits robustness to DoA estimation error representing the cumulative effect of uncertainties in urban environments and construction sites.