Modelling for sound annoyance evaluation of vehicle noise based on neural network

Based on the neural network (NN) technique, a new sound quality evaluation (SQE) model for synthetical annoyance of vehicle noises is proposed in this paper. Based on the measured noise signals of a sample vehicle, the psychoacoustical models for loudness, sharpness and roughness are investigated. Annoyance values of the measured vehicle noises are estimated by a jury test following an anchor-scaled scoring method. A NN-based model for annoyance evaluation of vehicle noise, so-called VNA-NNM, is carried out. Finally, the VNA-NNM model is verified by the leave-one-out algorithm. The results suggest that the VNA-NNM model is accurate and effective for estimating sound quality of vehicle interior noises, which is instructive and useful for vehicle acoustical designs. Streszczenie. Zaproponowano model oceny jakości dźwięku i rozdrażnienia dźwiękiem szumu samochodowego. Model bazuje na sieciach neuronowych i wykorzystuje zmierzony sygnał szumu samochodowego oraz model psychoakustyczny. (Modelowanie dźwięków drażniących pochodzących od pojazdów mechanicznych z wykorzystaniem sieci neuronowych)