Considering the well-known features of the noise emitted by moving sources, a number of vehicle characteristics such as speed, unladen mass, engine size, year of registration, power and fuel were recorded in a dedicat...Considering the well-known features of the noise emitted by moving sources, a number of vehicle characteristics such as speed, unladen mass, engine size, year of registration, power and fuel were recorded in a dedicated monitoring campaign performed in three 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 sta- tistically relevant features. In order to compare the vehicle transit noise in different environmental condition, all 1/3-octave band spectra were normalized and analysed. Un- supervised clustering algorithms were employed to group together spectrum levels with similar profiles. Our results corroborate the well-known fact that speed is the most rele- vant characteristic to discriminate between different vehicle noise spectrum. In keeping with this fact, we present a new approach to predict analytically noise spectra for a given vehicle speed. A set of speed-dependent analytical functions are suggested in order to fit the normalized average spectrum profile at different speeds. This approach can be useful for predicting vehicle speed based purely on its noise spectrum pattern. The present work is complementary to the accurate analysis of noise sources based on the beamforming technique.展开更多
文摘Considering the well-known features of the noise emitted by moving sources, a number of vehicle characteristics such as speed, unladen mass, engine size, year of registration, power and fuel were recorded in a dedicated monitoring campaign performed in three 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 sta- tistically relevant features. In order to compare the vehicle transit noise in different environmental condition, all 1/3-octave band spectra were normalized and analysed. Un- supervised clustering algorithms were employed to group together spectrum levels with similar profiles. Our results corroborate the well-known fact that speed is the most rele- vant characteristic to discriminate between different vehicle noise spectrum. In keeping with this fact, we present a new approach to predict analytically noise spectra for a given vehicle speed. A set of speed-dependent analytical functions are suggested in order to fit the normalized average spectrum profile at different speeds. This approach can be useful for predicting vehicle speed based purely on its noise spectrum pattern. The present work is complementary to the accurate analysis of noise sources based on the beamforming technique.
