Abstract: In order to establish a multi-dimensional evaluation model for the interior sound quality of a domestic sport utility vehicle (SUV), steady-state and non-stationary noise sample signals are first collected from the driver's position in the car, and a noise evaluation review panel is organized. Through research, the user profile of the reviewer is determined, and sound quality is divided into two dimensions: comfort and sportiness. Subjective evaluations of sound quality under different dimensions are conducted using a grading method. Then, based on Matlab, the objective parameters of the sound quality of the sample signals are calculated and correlated with the subjective evaluation scores to determine the parameters strongly correlated with comfort and sportiness. Finally, the sample signals are divided into training samples and validation samples, and the objective parameters and subjective evaluation scores of the training samples are used as input and output parameters for the objective evaluation model of sound quality. A genetic algorithm-back propagation (GA-BP) Neural Network mathematical model is established using Matlab Simulink. Through verification of the samples, the GA-BP model has the smallest error in evaluating sound quality in both dimensions, with average prediction errors of 3.63% and 3.04%, respectively.