Statistical Evaluation and Regression Analysis of Vehicle Sound Quality

A statistical approach to evaluate the subjective perception of the annoyance caused by the vehicle noise was presented in this paper. After recording the noises of Sanfeng, Huali and Xiali at speeds of 30, 40, 50, 60, 70 and 80 km/h respectively, the annoyance of the vehicle noises was evaluated in the testing room using paired comparison method, and the sound quality metrics and subjective annoyance were then distilled. Loudness, sharpness, roughness, periodicity and impulsiveness were selected for each of the vehicle noises. By correlation analysis method, it can be found that loudness has a higher correlation (0.91) with annoyance than other parameters. Meanwhile, sharpness, periodicity, roughness and impulsiveness have correlation with subjective perception with correlation coefficients being 0.84,-0.82, 0.62 and 0.87, respectively. The result of multiple regression analysis shows that calculated annoyance obtained by the regression equation can explain the perceptual annoyance and the regressed evaluation model is feasible to evaluate the sound quality of vehicle.