Research On Subjective Evaluation Prediction Model Of Passenger Vehicles Door Closing Sound Quality

With the continuous iterative updating of technology and the continuous improvement of people’s living standards,the number of vehicles in China is increasing,and people have put forward higher requirements for the driving comfort of passenger vehicles.As one of the "three major public hazards",improving sound quality has become an important research topic.The sound quality when closing the door directly affects the user’s driving experience.Good sound quality can bring happiness and satisfaction to the user.Therefore,the research on the sound quality of closing the door of passenger vehicles is of great significance and value.Based on the sound signal acquisition test received by drivers and passengers during the door closing process of17 different passenger vehicles,this dissertation carried out the research on the subjective evaluation and prediction model of the sound quality of the door closing sound of the car,that is,the quality of the door closing sound of the car can be quickly predicted through the collected data of the door closing sound,which can not only effectively and accurately evaluate the quality of the door closing sound,but also save the time cost of the subjective evaluation test.The research contents mainly include the following aspects:Firstly,this dissertation conducted an objective evaluation test on the sound quality of passenger vehicle door closing.First of all,the door closing sound signals of 17 passenger vehicles of different models,grades and brands are collected.The measuring points include the signals received by the driver and passenger’s ear when the door is closed(the outer door of the driver’s car and the right ear of the passenger’s car).The purpose is to focus on the impact of closing the door on the driver and passenger at the same time.A total of 34 sound samples were obtained through screening: 17 at the outer door of the first type of driver’s car and 17 at the right ear of the second type of passenger’s car;Then select the objective evaluation parameters,including A-weighted sound pressure level,loudness,sharpness,roughness,and volatility,and extract the objective evaluation parameters for each sound sample to complete the objective evaluation test of door closing sound quality.Secondly,this dissertation conducted a subjective evaluation test on the sound quality of passenger vehicle door closing.The preparation of the subjective evaluation test includes the selection of the subjective evaluation index,the subjective evaluation method,the evaluation crowd and the environment.Then the evaluators use the collected samples of the door closing sound to conduct the subjective evaluation test and obtain the subjective evaluation results of each sample.At the same time,the subjective evaluation test data are processed to eliminate the invalid or large error data.The processing methods include the same sample pair evaluation,exchange sample pair evaluation,triangle cycle judgment and weight miscalculation analysis.Finally,the evaluation results of 21 evaluators are retained to obtain the subjective evaluation value.Thirdly,the subjective evaluation model of door closing sound quality is established by using various prediction algorithms.The objective evaluation test result is used as the input of the model,and the subjective evaluation result is used as the output of the model to establish the door closing sound quality prediction model.Among them,the research on prediction algorithm includes multiple linear regression analysis,BP neural network algorithm,BP neural network algorithm based on genetic algorithm(GA-BP)and BP neural network algorithm based on particle swarm optimization(PSO-BP).The test data is divided into training samples and prediction samples,which are brought into the model for training and prediction respectively.The prediction accuracy of the prediction model is evaluated by performance indicators(mean square error,average relative error and determination coefficient),and the final result is that the GA-BP neural network algorithm has the highest prediction accuracy and the multivariate linear regression analysis is the worst.Finally,based on feature extraction,the subjective evaluation and prediction model of door closing sound quality is established.Based on the GA-BP neural network algorithm obtained above,the objective evaluation parameters are discarded as the input of the model.Instead,the non-stationary signal processing methods of empirical mode decomposition(EMD)and wavelet packet transform(WPT)are used to process the door closing sound signal respectively,and feature extraction is performed.The feature vector is used as the input of the model for training and prediction,and the prediction performance of GA-BP,EMD-GA-BP and WPT-GA-BP is compared,Finally,the WPT-GA-BP neural network algorithm model can predict the subjective evaluation of door closing sound quality most accurately.