The evaluation of vehicle interior impact noise inducing by speed bumps based on multi-features combination and support vector machine

Abstract In this paper, a multi-features combination process was proposed to evaluate the vehicle impact noise inducing by speed bumps. Vehicle road tests were carried out for four vehicles. The test vehicles were driven at different speeds on roads with two types of speed bumps. The impact noises of the vehicle were recorded in the interior cabin to build up the impact noise database. Empirical mode decomposition method was adopted to break up the signal, and Teager-Kaiser energy operator method was employed to analyze the fast change of the energy. Based on these methods, the acoustic features and non-acoustic features were extracted from the noise samples. To compress the eigenvector of the dataset, the principal component analysis method was also used in this study. Jury test was conducted to evaluate the characteristic of the noise samples. The intelligent predicting method-support vector machine was used to build up the acoustic predicting model of the impulsiveness for the impact noise. The combined multi-features input dataset exhibits better performance for the predicting model. These results indicate that the combined multi-features of the noise samples can be used to estimate the impulsiveness of the impact noise without jury evaluation.