Acoustic Detection of Moving Vehicles

Abstract

This chapter outlines a robust algorithm to detect the arrival of a vehicle of arbitrary type when other noises are present. It is done via analysis of its acoustic signature against an existing database of recorded and processed acoustic signals. To achieve it with minimum number of false alarms, a construction of a training database of acoustic signatures of signals emitted by vehicles using the distribution of the energies among blocks of wavelet packet coefficients (waveband spectra, see Sect. 4.6) is combined with a procedure of random search for a near-optimal footprint (RSNOFP). The number of false alarms in the detection is minimized even under severe conditions such as: signals emitted by vehicles of different types differ from each other, whereas the set of non-vehicle recordings (the training database) contains signals emitted by planes, helicopters, wind, speech, steps etc. The described algorithm is robust even when the tested conditions are completely different from the conditions where the training signals were recorded. This technique has many algorithmic variations. For example, it can be used to distinguish among different types of vehicles. The described algorithm is a generic solution for process control that is based on a learning phase (training) followed by an automatic real-time detection.