Directional spectrum estimation of moving vehicles

The simulation of urban environments requires knowledge of vehicle directional noise emission characteristics to increase accuracy and provide more authenticity to 3D audio. This paper presents a methodology to extract the directional sound energy pattern emitted by a moving vehicle using a simple microphone array. The method involves recording the vehicle's pass-by sound with a set of two horizontal sensors and, based on the spectrogram and a modified direction of arrival (DoA) algorithm, determining the energy spectrum according to the direction between the source and the receivers. The proposed method is initially tested with simulated data, where the analysis signals are generated by simulating the passage of a sound source with a known directivity pattern. Using white noise as the source signal, the proposed algorithm successfully recovered the directional pattern with errors of up to 1 dB. The method is then applied to real measurements conducted in a controlled environment, where four vehicles passed by the sensors at different speeds. As there was no ground truth for evaluation, the spectral characteristics of the directional data were compared with similar data in the literature. The results showed an agreement between the estimated and expected data, indicating that the proposed method is capable of extracting horizontal directional information from passing vehicles. This can be used to determine the general directional emission patterns of different vehicle categories or models for use in acoustic virtual environments.