Research On Sound Quality Evaluation Of Vehicle Interior Noise Using Human-ear Physiological Model

With the increase of automobile penetration rate,automobiles have become the first source of urban noise pollution,and the noise control of automobiles is of great significance to ensure people’s quality of life.However,the early studies on automotive noise control mainly focused on physical parameters such as sound pressure level,ignoring the real feelings of people.As consumers have higher and higher requirements for car comfort,the sound quality,which reflects people’s subjective auditory perception of noise,has attracted more and more attention from manufacturers and consumers.Rapid and accurate evaluation of the sound quality of interior noise has become the key to the acoustic design of modern automobiles.Sound quality evaluation can be divided into two types: subjective evaluation and objective evaluation.Among them,subjective evaluation cannot be directly used in the acoustic design of automobiles due to its high time cost;and objective evaluation has the problem of low prediction accuracy.Therefore,it is of great significance to establish a high-accuracy objective sound quality prediction model for automotive acoustic design.In objective evaluation,finding some better objective indicators to evaluate sound quality will improve the accuracy of sound quality prediction.Traditional psychoacoustic indicators mainly approximate the perception characteristics of the human ear,but do not have sound transmission characteristics based on the physiological and anatomical structure of the real human ear.Meanwhile,due to the complex geometry,ultrastructure,and complex physiological structure of the human ear,it is difficult to establish a complete model with the physiological structure of the human ear.The purpose of this dissertation is to establish a more accurate evaluation model for sound quality of the interior noise.It is necessary to calculate the psychoacoustic parameters of interior noise in order to improve the accuracy of the sound quality evaluation model for interior noise.Therefore,this dissertation proposes a calculation model for psychoacoustic parameters(physiological loudness,physiological sharpness,physiological roughness)based on the human ear physiological model.To enable the proposed psychoacoustic parameters to more accurately represent human auditory perception,it is necessary to establish an accurate ear model for processing sound signals.Therefore,establishing a high-precision physiological model of the human ear is the foundation for the subsequent establishment of an evaluation model for interior noise quality,and is an important research object of this article.This article describes the following structure for the content of the dissertation: firstly,it is necessary to accurately model the human ear physiological model,including establishing a finite element model of the human ear and an active cochlear model,to provide a human ear physiological model for the subsequent calculation of psychoacoustic parameters(physiological loudness,physiological sharpness,physiological roughness);Secondly,through interior noise collection experiment,establish a noise sample library and subjectively evaluate it;Finally,a new sound quality evaluation model is constructed by combining physiological loudness,physiological sharpness,physiological roughness,and subjective annoyance of interior noise.It provides a new method and evaluation indicator for the sound quality evaluation of vehicle interior noise,which has important theoretical significance and engineering application value.The main contents are as follows:(1)The interior noises of a class B fuel car were collected.USB-4431 data acquisition card was used to connect the microphones with computers and other devices,and Labview was used to design a data acquisition program to collect and store noise data in the vehicle.The noises were collected and stored at different conditions,speeds,and locations within the vehicle.The interior noises were subjectively evaluated by the subjective evaluation experiment.The subjective evaluation index is annoyance,and the subjective evaluation experiment was conducted in a double-walled sound attenuating chamber.The Sennheiser HD650 high-fidelity headset was used as sound playback devices,and 28 evaluators used the adaptive grouped paired comparison method to score the noise.The Python programs were used to analyze the effectiveness and consistency of several misjudgment methods of all evaluators.Finally,three evaluators’ unstable evaluation results were removed,and the average of the evaluation results of the remaining 25 evaluators was used as the subjective evaluation value.(2)Based on the real μCT image of the temporal bone,considering the real threedimensional structure of the human ear,a three-dimensional model of the human ear including the ear canal,middle ear and spiral cochlea was constructed by using reverse molding technology.The finite element model of the human ear was developed based on the complex physiological structure of the human ear and considering the complex nonlinear structure of the human ear.Among them,the soft tissue material of the human ear,cochlear basilar membrane,and ossicles were modeled using viscoelastic materials,anisotropic materials,and linear elastic materials,respectively.The acoustic-solid coupling was established for the three states of matter in the human ear(gas,solid,and liquid)by the multi-field coupling modeling method.Considering the influence of the presence of overflowable pores in the cochlea,a cochlear model including a third window was established.By analyzing the responses of the outer ear,the middle ear,and the cochlea,the accuracy of the model was verified,and the acoustic characteristics of the human ear can be simulated by using the physiological model of the human ear.(3)Considering the activity of the cochlea,according to the basic theory of the active amplification mechanism of the cochlea and the feed-forward mechanism of the basilar membrane,combined with the passive finite element model of the cochlea,an active finite element model of the cochlea in the frequency domain was established.According to the basilar membrane feed-forward mechanism,the active force of the hair cells on the basilar membrane is calculated from the passive cochlear response,combined with the nonlinear gain of the basilar membrane,Abaqus was used to call the Python language program,and the passive cochlea was iterated repeatedly to solve the active cochlear response.The finite element model of the active cochlea takes a long time to calculate and is only applicable to the frequency domain,so a nonlinear timedomain cochlear transmission line model was introduced.By calculating the motion,nonlinear compression characteristics and impulse response of the active cochlea,the amplification function and nonlinear compression of the active cochlea were analyzed,which provides a theoretical basis and method for in-depth understanding of the sound transmission characteristics of the human ear.(4)Based on the finite element model of the outer and middle ears and the transmission line model of the cochlea,a bidirectional coupling model of the human ear finite element-transmission line was established.The sound signal passes through the outer and middle ear models,and drives the cochlea through the stapes footplate.Furthermore,the cochlear input impedance of the cochlear hinders the movement of the stapes,and acts on the stapes footplate.By comparing the middle ear and cochlea responses of the human ear finite element-transmission line coupling model with relevant experimental data,it was verified that the model can simulate the sound transmission characteristics of the physiological anatomical structure of the human ear.Based on the velocity response of the basilar membrane of the human ear physiological model,a loudness perception model related to nerve firing was established.In addition,according to the velocity response of the basilar membrane in the human ear physiological model,the physiological excitation level of sharpness and the physiological excitation difference of roughness were calculated.The physical response of the basilar membrane was converted into sharpness perception and roughness perception in the subjective auditory sensation.(5)The loudness,sharpness,and roughness perception model established based on the physiological model of the human ear were used to calculate the physiological loudness,sharpness,and roughness of the collected vehicle interior noise.The correlation coefficients of calculated physiological loudness,sharpness,roughness and subjective evaluation values are 0.936,0.859,0.839,respectively,which are higher than the correlation coefficients of Moore loudness/Zwicker loudness,sharpness,roughness calculated by traditional methods and subjective evaluation values.It showed the superiority of the three psychoacoustic parameters calculated based on the physiological model in the evaluation of the sound quality of vehicle interior noise.Taking physiological loudness,sharpness,and roughness as input and subjective evaluation results as target,two types of sound quality model of the interior noise were established by BP neural network and Tab Net model.In the BP neural network sound quality model,the model was optimized by the sparrow search algorithm,and the prediction error is 4.42%,which is superior to the loudness,sharpness,and roughness calculated by traditional methods as inputs.In addition,a Tab Net model was used to predict the sound quality of interior noise.The results show that the prediction error of the Tab Net sound quality model is 3.58%,which has a lower prediction error compared to the optimized BP neural network sound quality model.This dissertation has 85 figures,12 tables and 174 references.