关键词: 电动汽车；声品质；智能化；回归分析；神经网络

Abstract: Electric vehicles feature a low overall noise level yet complex noise composition. A single sound pressure level fails to accurately match human auditory perception, so subjective and objective evaluation is introduced for electric vehicle sound quality research. However, subjective evaluation is time-consuming and labor-intensive with a shortage of professional personnel, making it urgent to realize accurate subjective evaluation prediction through objective parameters. This paper conducts relevant research accordingly. It first analyzes the sound source characteristics of electric vehicles, investigates the subjective and objective evaluation methods for electric vehicle sound quality, establishes an intelligent correlation model for the subjective and objective evaluation of electric vehicle sound quality, and implements the whole-process practical application. The research results show that the regression analysis model has a simple prediction process and fast calculation speed. The neural network model imposes no strict requirements on input parameters and achieves higher prediction accuracy than the linear regression model, with the model prediction error within 0.4. The high-precision prediction practice provides a favorable reference for the development of a convenient, efficient and low-cost approach for the intelligent evaluation of automobile noise.

Key words: electric vehicles; sound quality; intelligence; regression analysis; neural network