Study On Subjective And Objective Evaluation Model For Psychoacoustical Quality Of Vehicle Interior Noise

With the improvement of people’s living quality, the problem of noise pollution hasbeen catching the pulic concerns. The interior noise qulity is one of the important parts inthe sound qulity evaluation (SQE) engineering of vehicles. The subjective and objectiveevaluation models for psychoacoustical quality of vehicle interior noise are developedbased on the studies of former research in this dissertation.According to the automotive interior noise standard named GB/T18697, firstly, someinterior noises under different working conditions of a sample car are measured. And theacquired noise signals are pre-processed and saved in a noise database. A new improvedacoustical roughness model is proposed by optimizing algorithm in the existing roughnessmodels. Referring to some related standards, the calculation programs for pychoacousticalparameters, such as loudness, sharpness and roughness, are compiled based on the Matlabsoftware. Based on a comparison of the subjective evaluation methods, a jury is organized;and the method so-called optimized semantic differential with reference signals is used foracoustical subjective evaluation of the noises in the databasetests. The result of subjectiveevaluation is validated by a set of tests. A SQE model is established by using BP neuralnetwork, in which the calculated psychoacoustical parameters and jury test results are setas the inputs and outputs of the network, respectively. The network model is optimized bymodifying the structure and weights. Finally, the prediction accuracy of the newly designedSQE model is validated and analyzed by experimental results.Based on the above findings, the calculation results of the self-made roughness modelare in accord with those from the jury tests, which suggests that the optimized roughnessalgorithm presented in this dissertation has enough accuracy. The SQE results predicted bythe well-trained neural network model agree with the subjective evaluation results.Therefore, a conclusion can be drawn that the developed SQE network is correct and effective for vehicle interior noises. The proposed SQE network may be directly used forsound quality analysis of vehicle interior noises, which is beneficial for acoustical designand improvement in vehicle engineering.