Sound quality prediction of unsteady vehicle interior sound

Abstract

In this paper, the noise samples of pure electric vehicles under unsteady-state are collected. Then, the subjective evaluation test is carried out after pre-processing such as screening and intercepting. The subjective sound quality scores of the noise samples are obtained. Meanwhile, the noise samples are calculated of four conventional psychoacoustic objective parameters such as loudness, sharpness, roughness, and speech intelligibility. Preprocessing by ensemble empirical mode decomposition (EEMD) is performed. Characteristic parameters of noise samples such as time-frequency domain fractal dimension difference and sample entropy of noise samples are obtained based on fractal dimension and sample entropy theory. Finally, the quality prediction model of vehicle interior sound is established based on six characteristic parameters and BP neural network. The results show that the prediction effect is excellent for the subjective scores of sound quality of unsteady interior sound in the vehicle.