Sound Quality Evaluation of the Interior Noise of Pure Electric Vehicle Based on Neural Network Model

Abstract

Based on neural network model, a method for quantitative sound quality (SQ) evaluation of the interior noise of a pure electric vehicle (PEV) is presented in this paper. The method can be divided into four steps. First, the interior noises under different speeds of a PEV are collected through the interior noise test of the PEV. Subsequently, one physical acoustic parameter (A-weighted sound pressure level) and six psychoacoustic parameters (loudness, fluctuation strength, tonality, roughness, articulation index, and sharpness) are applied to describe the noise samples for objective evaluation of SQ. In the third step, five semantic evaluation indexes, namely, “annoying or pleasing,” “harsh or sweet,” “weak or powerful,” “promiscuous and pure,” and “unobservable or perceptible,” are proposed based on semantic differential method, which are used for subjective evaluation of SQ by jury tests. Finally, the neural network model for SQ evaluation of the interior noise of the PEV is established, the SQ characteristics of the interior noise of the PEV are evaluated, as well as revealing the coefficient weight of influencing factors. This model can be used for SQ prediction and evaluation of the interior noise of the PEV considering that the average error is 9%.