Traffic acoustical classes based upon vehicle characteristics

The identification of sound events and their recognition is a recurrent theme in outdoors noise monitoring. A sound generating device is generally identified according to its acoustic features which are referred to as acoustic signatures and are usually employed to discriminate among vehicles. Considering the well-known features of the noise emitted by moving vehicles, an experiment has been planned in order to obtain detailed information on the moving sources. The experiment has been performed in 3 different places, each characterized by different number of lanes and the presence of nearby reflective surfaces. A full database of 144 vehicles (cars) was used to identify statistically relevant features. In order to compare vehicle transit noise in different environmental condition, all 1/3 octave band spectra were normalized and statistically analysed. This approach allowed vehicle spectral signatures to be organized in acoustical classes based upon vehicle characteristics (speed, fuel, age, etc.). The optimal number of traffic acoustical classes came from the compromise between satisfactory discrimination and the need to limit the number of groups.