A bimodal sound source model for vehicle tracking in traffic monitoring

The paper addresses road traffic monitoring using a compact microphone array. More precisely, estimation of both speed and wheelbase distance of detected vehicles is performed. The detection algorithm is based on the comparison between theoretical and measured correlation time series using the two dimensional Bravais-Pearson correlation coefficient. The tracking step is conducted with a particle filter specifically designed to model the position-variant bimodal sound source nature of the vehicles, i.e. taking into account the sound emitted by both vehicle axles. Sensitivity and performance studies using simulations and real measurements show that the bimodal approach reduces the tracking failure risk in harsh conditions when vehicles are tracked, at the same time, in opposite directions.