Design of the Evaluation Model of Acceleration Sound Quality of Automobile Based on MLP

<div class="section abstract"><div class="htmlview paragraph">The acceleration vibe of a car's engine can be enhanced and a brand-specific auditory identity can be created via active sound design. Currently, experienced engineers are desperately required when the active sound design for car acceleration roar was processing, which consumed substantial time and human resources. Therefore, it is critical to conduct a research on the evaluation model for estimating car acceleration sound quality to improve sound design efficiency and reducing costs. 1,003 acceleration roars samples of common cars were collected in this paper, all of which could be commonly heard by the road. Nine psychoacoustic objective parameters, such as loudness, sharpness, and roughness, were calculated through Artemis Suite software, establishing a database for the sound quality of car acceleration sounds.Moreover, subjective evaluations of sound playback and objective data analysis were conducted to obtain the ratings of acceleration sounds. Firstly, five objective psychoacoustic parameters are constructed to evaluate the vehicle sound quality based on the correlation analysis. Subsequently, an evaluation and prediction model of acceleration sound quality based on MLP(Multi layer Perceptron) was constructed, with psychoacoustic parameters as input features and subjective evaluation scores as output, which correlates objective test data with subjective evaluation results. This model was determined as the evaluation model for car acceleration sound quality for this case, and the average prediction accuracy rate of which is more than 90%, providing key technical support for promising active sound design of car engines.</div></div>