In-vehicle noise evaluation and prediction implementation method, device, equipment and medium

Abstract

The invention discloses an in-vehicle noise evaluation and prediction implementation method, device, equipment and medium, and relates to the technical field of vehicle application, the method comprises the following steps: carrying out in-vehicle noise test on a target vehicle to obtain in-vehicle noise spectrum data, and obtaining a subjective evaluation result based on a subjective evaluation mode; carrying out frequency band division on the in-vehicle noise spectrum data, and labeling by adopting a subjective evaluation result corresponding to the in-vehicle noise spectrum data to obtain a data set so as to realize construction of a training set and a verification set; training the constructed deep learning network model by using the training set, and verifying the trained deep learning network model by using the verification set; and according to the verification result, based on a deep learning network model, performing subjective evaluation result prediction of the in-vehicle noise, or optimizing frequency band division of the in-vehicle noise spectrum data to obtain a training set again, and training the deep learning network model again. According to the invention, accurate prediction of subjective evaluation of the in-vehicle noise can be realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of vehicle application, in particular to an in-vehicle noise evaluation and prediction implementation method, device, equipment and medium. BACKGROUND

[0002] With the improvement of people's living standards, the number of private cars is also increasing year by year, and the performance of automobile NVH (Noise, Vibration, Harshness, noise, vibration and harshness) is more and more valued by market users. The evaluation of vehicle noise level generally adopts objective test and subjective evaluation. Objective test is to collect and analyze in-vehicle noise data to obtain noise spectrum data; subjective evaluation is that professional personnel subjectively score the vehicle noise level. The good or bad of the vehicle noise level is ultimately evaluated by the user.

[0003] If the subjective feeling of customers on the vehicle noise level can be accurately predicted during the vehicle development stage, it can guide the identification of NVH problems and the verification of schemes, which is of great significance to the development of vehicle NVH. Since there is not a linear corresponding relationship between the test data of sound and subjective feeling, only relying on the comparison of the size of the test data sound pressure to predict the subjective feeling of customers on the in-vehicle noise is not very accurate. At present, the industry can only make auxiliary judgments through the size of noise sound pressure and some sound quality parameters, and there is no mature way to predict the subjective evaluation result of in-vehicle noise based on test data. SUMMARY

[0004] The present application provides an in-vehicle noise evaluation and prediction implementation method, device, equipment and medium, which can realize accurate prediction of in-vehicle noise subjective evaluation.

[0005] In a first aspect, the embodiments of the present application provide an in-vehicle noise evaluation and prediction implementation method, which comprises: Performing in-vehicle noise test on a target vehicle to obtain in-vehicle noise spectrum data, and obtaining subjective evaluation results based on a subjective evaluation method; Dividing the in-vehicle noise spectrum data into frequency bands, and using the subjective evaluation results corresponding to the in-vehicle noise spectrum data to label to obtain a data set, so as to construct a training set and a validation set; Using the training set to train a constructed deep learning network model, and using the validation set to verify the deep learning network model after training; According to the verification result, predicting the subjective evaluation result of in-vehicle noise based on the deep learning network model, or optimizing the frequency band division of in-vehicle noise spectrum data to obtain a training set again, and training the deep learning network model again.

[0006] In combination with the first aspect, in an implementation, the in-vehicle noise test on the target vehicle obtains in-vehicle noise spectrum data, and a subjective evaluation result is obtained based on a subjective evaluation method, and specifically includes: determining a test frequency range of the in-vehicle noise test, obtaining a plurality of noise test data by performing the in-vehicle noise test on the target vehicle, and performing spectrum analysis on each of the noise test data to obtain in-vehicle noise spectrum data; performing subjective evaluation on the in-vehicle noise while obtaining the noise test data by performing the in-vehicle noise test, and obtaining a subjective evaluation result corresponding to the in-vehicle noise spectrum data.

[0007] In combination with the first aspect, in an implementation, the subjective evaluation result prediction of the in-vehicle noise is performed based on the deep learning network model according to the verification result, or the frequency band division of the in-vehicle noise spectrum data is optimized to obtain a training set again, and the deep learning network model is trained again, and specifically includes: obtaining the RMES value and the R2 value of the verification result of the trained deep learning network model according to the verification set: if the RMES value is greater than a set value or the R2 value is less than a preset value, the frequency band division of the in-vehicle noise spectrum data is optimized, a data set is obtained again to train the deep learning network model, and then the subjective evaluation result prediction of the in-vehicle noise is performed based on the deep learning network model trained again; if the RMES value is not greater than the set value and the R2 value is not less than the preset value, the deep learning network model is used to predict the subjective evaluation result of the in-vehicle noise.

[0008] In combination with the first aspect, in an implementation, the in-vehicle noise spectrum data is divided into frequency bands, and the subjective evaluation result corresponding to the in-vehicle noise spectrum data is labeled to obtain a data set, so as to construct a training set and a verification set, and specifically includes: the in-vehicle noise spectrum data is divided into frequency bands based on a 1 / 3 octave division method, and each in-vehicle noise spectrum data is divided into frequency bands; a 32-dimensional vector is constructed as a test data based on the amplitude of each frequency band of the current in-vehicle noise spectrum data, and the label of the current test data is the subjective evaluation result corresponding to the current in-vehicle noise spectrum data, a plurality of test data are obtained to obtain a data set; the data set is divided according to a preset proportion to construct a training set and a verification set.

[0009] In combination with the first aspect, in an implementation, the optimization of the frequency band division of the in-vehicle noise spectrum data specifically includes: For 32 frequency bands of 1 / 3 octave division, a preset number of frequency bands are randomly selected each time, the fl value of each selected frequency band is updated based on a first preset algorithm, the fu value of each selected frequency band is updated based on a first set algorithm or a second set algorithm, a new frequency band division mode is obtained, and finally a plurality of new frequency band division modes are obtained; Based on each obtained new frequency band division mode, all in-vehicle noise spectrum data are divided into frequency bands, and input into the trained deep learning network model to obtain a prediction result, and the new frequency band division mode corresponding to the highest accuracy is taken as the final frequency band division mode. The in-vehicle noise spectrum data is divided into frequency bands using the final frequency band division mode, and a data set is obtained again. If the first frequency band of the 32 frequency bands is included in the selected frequency bands, the fl value of the first frequency band is not updated, and if the last frequency band of the 32 frequency bands is included in the selected frequency bands, the fu value of the last frequency band is not updated.

[0010] In combination with the first aspect, in an implementation manner, The first preset algorithm is fla(i)=fu(i-1), where fla(i) represents the updated fl value of the i-th frequency band, and fu(i-1) represents the fu value of the (i-1)-th frequency band. The first set algorithm is fua(i)=fu(i)-(fu(i)-fl(i))×0.1, where fua(i) represents the updated fu value of the i-th frequency band, fu(i) represents the fu value of the i-th frequency band, and fl(i) represents the fl value of the i-th frequency band. The second set algorithm is fua(i)=fu(i)+(fu(i)-fl(i))×0.1.

[0011] In combination with the first aspect, in an implementation manner, The deep learning network model is a DNN, and the nonlinear activation function of the DNN is selected as Relu. When training the deep learning network model, the optimizer is selected as the stochastic gradient descent method, and the loss function is selected as L2-Loss.

[0012] Secondly, the present application provides an in-vehicle noise evaluation and prediction implementation device, which comprises: The acquisition module is configured to perform in-vehicle noise testing on the target vehicle to obtain in-vehicle noise spectrum data, and obtain a subjective evaluation result based on a subjective evaluation method. a construction module configured to divide the in-vehicle noise spectrum data into frequency bands, and label the in-vehicle noise spectrum data with subjective evaluation results corresponding to the in-vehicle noise spectrum data to obtain a data set, so as to construct a training set and a verification set; a training module configured to train a constructed deep learning network model using the training set, and verify the deep learning network model trained using the verification set; an execution module configured to, according to the verification result, predict the subjective evaluation result of the in-vehicle noise based on the deep learning network model, or optimize the frequency band division of the in-vehicle noise spectrum data to obtain a training set again, and train the deep learning network model again.

[0013] In a third aspect, an in-vehicle noise evaluation and prediction implementation device is provided, which comprises a processor, a memory, and an in-vehicle noise evaluation and prediction implementation program stored in the memory and executable by the processor. When the in-vehicle noise evaluation and prediction implementation program is executed by the processor, the steps of the in-vehicle noise evaluation and prediction implementation method described above are implemented.

[0014] In a fourth aspect, a computer readable storage medium is provided, which stores an in-vehicle noise evaluation and prediction implementation program. When the in-vehicle noise evaluation and prediction implementation program is executed by a processor, the steps of the in-vehicle noise evaluation and prediction implementation method described above are implemented.

[0015] The technical scheme provided by the embodiments of the present application has the following beneficial effects: By dividing the in-vehicle noise spectrum data into frequency bands, and labeling the in-vehicle noise spectrum data with subjective evaluation results corresponding to the in-vehicle noise spectrum data to obtain a data set, so as to construct a training set and a verification set, then training a constructed deep learning network model using the training set, and verifying the deep learning network model trained using the verification set, then according to the verification result, predicting the subjective evaluation result of the in-vehicle noise based on the deep learning network model, or optimizing the frequency band division of the in-vehicle noise spectrum data to obtain a training set again, and training the deep learning network model again, the in-vehicle noise subjective evaluation can be accurately predicted based on the high-precision deep learning network model. BRIEF DESCRIPTION OF DRAWINGS

[0016] Figure 1 FIG. 1 is a flowchart of the in-vehicle noise evaluation and prediction implementation method of the present application; Figure 2 FIG. 2 is a schematic diagram of the functional modules of the in-vehicle noise evaluation and prediction implementation device of the present application; Figure 3 FIG. 3 is a schematic diagram of the hardware structure of the in-vehicle noise evaluation and prediction implementation device of the present application. DETAILED DESCRIPTION

[0017] In order to enable those skilled in the art to better understand the present invention, the following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0018] In order to make the objectives, technical solutions and advantages of this application clearer, the implementation methods of this application will be further described in detail below with reference to the accompanying drawings.

[0019] First, an embodiment of the present application provides a method for implementing in-vehicle noise evaluation prediction, which uses a deep neural network algorithm to accurately predict the subjective evaluation results of in-vehicle noise through the spectrum of in-vehicle noise, and improves the model prediction accuracy by optimizing the frequency band division method. It can be applied to the data stage of vehicle development. Before the actual vehicle is produced, the spectrum of in-vehicle noise can be calculated through CAE simulation and then the present application can be used to predict the results of customers' subjective evaluation of in-vehicle noise, thereby identifying risks in advance; and after the model training is completed, it is only necessary to use the in-vehicle noise data tested during the vehicle development process to predict the subjective evaluation results of customers on in-vehicle noise.

[0020] In one embodiment, referring to Figure 1 , Figure 1 This is a flow chart of the method for implementing the in-vehicle noise evaluation and prediction in this application. Figure 1 As shown in FIG, the method for realizing the evaluation and prediction of the vehicle interior noise includes: S1: Perform interior noise testing on the target vehicle to obtain interior noise spectrum data, and obtain subjective evaluation results based on a subjective evaluation method; S2: Divide the in-car noise spectrum data into frequency bands and label them using the subjective evaluation results of the corresponding in-car noise spectrum data to obtain a dataset for the construction of training and validation sets. S3: Using the training set to train the constructed deep learning network model, and using the verification set to verify the trained deep learning network model; S4: Based on the verification results, the subjective evaluation results of the in-vehicle noise are predicted based on the deep learning network model, or the frequency band division of the in-vehicle noise spectrum data is optimized to obtain a training set again, and the deep learning network model is trained again.

[0021] Furthermore, in one embodiment, an interior noise test is performed on a target vehicle to obtain interior noise spectrum data, and a subjective evaluation result is obtained based on a subjective evaluation method, specifically including: S101: Determine the test frequency range of the in-vehicle noise test, perform in-vehicle noise test on the target vehicle to obtain a plurality of noise test data, and perform spectral analysis on each noise test data to obtain in-vehicle noise spectrum data; S102: While obtaining noise test data by in-vehicle noise test, subjectively evaluate the in-vehicle noise to obtain subjective evaluation results, and the subjective evaluation results correspond one-to-one to the in-vehicle noise spectrum data.

[0022] Specifically, first, a vehicle that needs to be evaluated for in-vehicle noise, i.e., a target vehicle, is selected, then the test frequency range of the in-vehicle noise test is determined, which is generally 0~25600Hz, then the target vehicle is tested for in-vehicle noise, and spectral analysis is performed to obtain in-vehicle noise spectrum data. In practical applications, each in-vehicle noise test obtains one noise test data, and subjective evaluation is performed to obtain subjective evaluation results, i.e., each in-vehicle noise test corresponds to in-vehicle noise spectrum data and subjective evaluation results, and the in-vehicle noise test is performed multiple times.

[0023] The subjective evaluation result is a subjective score of the in-vehicle noise test according to subjective feeling. The corresponding relationship between subjective feeling (performance evaluation) and score is shown in Table 1 below.

[0024] Table 1

[0025] Further, in an embodiment, the in-vehicle noise spectrum data is divided into frequency bands, and the subjective evaluation results corresponding to the in-vehicle noise spectrum data are labeled to obtain a data set, so as to realize the construction of the training set and the validation set, which specifically includes: S201: Divide the in-vehicle noise spectrum data into frequency bands based on the 1 / 3 octave division method to obtain the frequency bands of each in-vehicle noise spectrum data; S202: Construct a 32-dimensional vector as a test data based on the amplitude of each frequency band of the current in-vehicle noise spectrum data, and the label of the current test data is the subjective evaluation result corresponding to the current in-vehicle noise spectrum data, to obtain a plurality of test data to obtain a data set; S203: Divide the data set according to a predetermined proportion to construct a training set and a validation set.

[0026] Specifically, at the initial state, the single vehicle interior noise spectrum data is divided into 32 frequency bands by using 1 / 3 octave division method, and then the single test data in the data set is cleaned and labeled. The single test data is the amplitude of the frequency band of the single vehicle interior noise spectrum data, which is a 32-dimensional vector, and the label is the subjective evaluation result (subjective evaluation score) of the vehicle interior noise spectrum data. Then, the test data in the data set is randomly divided into a training set and a validation set, and the ratio of the training set to the validation set can be 7:3.

[0027] It should be noted that the deep learning network model of the present application is a DNN (a kind of deep neural network), and the output result is the subjective evaluation result of the vehicle interior noise (reflected in the form of score), and the nonlinear activation function of the DNN is Relu; when the deep learning network model is trained, the optimizer selects the stochastic gradient descent method, and the loss function selects L2-Loss.

[0028] Further, in an embodiment, based on the validation result, the subjective evaluation result of the vehicle interior noise is predicted based on the deep learning network model, or the frequency band division of the vehicle interior noise spectrum data is optimized to obtain the training set again, and the deep learning network model is trained again. Specifically, it comprises: According to the RMES value and R2 value of the validation result of the trained deep learning network model based on the validation set: If the RMES value is greater than the set value or the R2 value is less than the preset value, the frequency band division of the vehicle interior noise spectrum data is optimized, the data set is obtained again to train the deep learning network model, and then the subjective evaluation result of the vehicle interior noise is predicted based on the deep learning network model trained again. If the RMES value is not greater than the set value and the R2 value is not less than the preset value, the subjective evaluation result of the vehicle interior noise is predicted by using the deep learning network model. The RMES value and the R2 value are both network model evaluation indexes.

[0029] Specifically, after the training set is used to train the deep learning network model, the validation set is used to verify the model effect, and the model effect is confirmed based on the RMES value and the R2 value. If the RMES value is large or the R2 value is small, it indicates that the accuracy of the model is not enough, and the prediction accuracy of the model needs to be improved, that is, the frequency band division method of the vehicle interior noise spectrum data is optimized, the training set is obtained again, and the deep learning network model is trained using the training set obtained again to improve the prediction accuracy of the model. After that, the deep learning network model can be used to predict the subjective evaluation result of the vehicle interior noise. If the RMES value is not greater than the set value and the R2 value is not less than the preset value, it indicates that the prediction accuracy of the deep learning network model has met the requirements, and the deep learning network model can be used to predict the subjective evaluation result of the vehicle interior noise.

[0030] For the subjective evaluation result prediction of the in-vehicle noise in the present application, specifically, a prediction program is written using a deep learning network model, the in-vehicle noise spectrum data collected in the application stage is taken as input, and the predicted subjective evaluation result is output.

[0031] Further, in an embodiment, for the optimization of the in-vehicle noise spectrum data frequency band division, specifically: A1: For the 32 frequency bands of 1 / 3 octave division, a preset number of frequency bands are randomly selected each time, the fl value of each selected frequency band is updated based on the first preset algorithm, the fu value of each selected frequency band is updated based on the first or second preset algorithm, a new frequency band division mode is obtained, and finally a plurality of new frequency band division modes are obtained; A2: Based on each new frequency band division mode obtained, the frequency band division of all in-vehicle noise spectrum data is performed, and the prediction result is obtained by inputting into the trained deep learning network model, and the new frequency band division mode corresponding to the highest accuracy is taken as the final frequency band division mode; A3: The in-vehicle noise spectrum data is divided into frequency bands using the final frequency band division mode, and the data set is obtained again; Wherein, if the first frequency band of the 32 frequency bands is contained in the selected frequency bands, the fl value of the first frequency band is not updated, and if the last frequency band of the 32 frequency bands is contained in the selected frequency bands, the fu value of the last frequency band is not updated.

[0032] It should be noted that 1 / 3 octave division can be up to 32 frequency bands. For the frequency range of each frequency band in the 1 / 3 octave division mode, see Table 2 shown below.

[0033] Table 2

[0034] The in-vehicle noise spectrum data frequency band division optimization of the present application is to adjust the frequency range of part of the frequency bands in Table 2 above. For example, 10 frequency bands are randomly selected from the 32 frequency bands, the frequency range of each selected frequency band is adjusted, and the remaining 22 unadjusted frequency bands are combined to obtain a new frequency band division mode. For example, the 2nd, 3rd, 8th, 9th, 11th, 13th, 15th, 16th, 19th, and 20th frequency bands are randomly selected, then the frequency range of the 10 selected frequency bands is adjusted, and the frequency range of the remaining 22 unadjusted frequency bands is combined to obtain a new frequency band division mode. The 2nd, 4th, 8th, 9th, 11th, 13th, 15th, 17th, 20th, and 25th frequency bands are randomly selected, then the frequency range of the 10 selected frequency bands is adjusted, and the frequency range of the remaining 22 unadjusted frequency bands is combined to obtain a new frequency band division mode.

[0035] In the present application, the first preset algorithm is fla(i)=fu(i-1), where fla(i) represents the updated fl value of the i th frequency band, and fu(i-1) represents the fu value of the i-1 th frequency band; the first setting algorithm is fua(i)=fu(i)-(fu(i)-fl(i))x0.1, where fua(i) represents the updated fu value of the i th frequency band, fu(i) represents the fu value of the i th frequency band, and fl(i) represents the fl value of the i th frequency band; and the second setting algorithm is fua(i)=fu(i)+(fu(i)-fl(i))x0.1. For example, when the frequency range of the 2 nd frequency band is updated, fla(2)=fu(1)=17.5, and fua(2)=fu(2)-(fu(2)-fl(2))x0.1=22.1-(22.1-17.5)x0.1=21.64 or fua(2)=fu(2)+(fu(2)-fl(2))x0.1=22.56.

[0036] After multiple selection operations and updating of the selected frequency bands after each selection, multiple new frequency band division manners can be obtained. Using each new frequency band division manner to divide the in-vehicle noise spectrum data can construct a 32-dimensional vector, which is input into the deep learning network model to obtain a prediction result. The prediction result is compared with the subjective evaluation result corresponding to the in-vehicle noise spectrum data, and the accuracy is calculated. Thus, the accuracy corresponding to the current new frequency band division manner is obtained. The accuracy corresponding to each new frequency band division manner is obtained. The new frequency band division manner with the highest accuracy is used as the final frequency band division manner. Then, the in-vehicle noise spectrum data is divided into frequency bands using the final frequency band division manner. The data set is obtained again. The deep learning network model is trained again based on the obtained data set, so as to improve the prediction accuracy of the deep learning network model.

[0037] The in-vehicle noise evaluation prediction implementation method of the embodiments of the present application divides the in-vehicle noise spectrum data into frequency bands, and labels the subjective evaluation result corresponding to the in-vehicle noise spectrum data to obtain a data set, so as to construct a training set and a validation set. Then, the training set is used to train the constructed deep learning network model. The validation set is used to verify the trained deep learning network model. Then, based on the verification result, the subjective evaluation result of the in-vehicle noise is predicted based on the deep learning network model, or the frequency band division of the in-vehicle noise spectrum data is optimized to obtain the training set again, and the deep learning network model is trained again. Thus, the high-precision deep learning network model can be used to accurately predict the subjective evaluation of the in-vehicle noise.

[0038] In a second aspect, the embodiments of the present application also provide an in-vehicle noise evaluation prediction implementation device.

[0039] In an embodiment, refer to Figure 2 , Figure 2 FIG. 1 is a schematic diagram of a function module of an in-vehicle noise evaluation and prediction implementation device according to an embodiment of the present application. As shown in FIG. 1, the in-vehicle noise evaluation and prediction implementation device includes a collection module, a construction module, a training module, and an execution module. Figure 2

[0040] The collection module is configured to perform in-vehicle noise testing on a target vehicle to obtain in-vehicle noise spectrum data, and obtain subjective evaluation results based on a subjective evaluation method. The construction module is configured to divide the in-vehicle noise spectrum data into frequency bands, and label the in-vehicle noise spectrum data based on the subjective evaluation results to obtain a data set, so as to construct a training set and a validation set. The training module is configured to train a deep learning network model based on the training set, and validate the trained deep learning network model based on the validation set. The execution module is configured to predict the subjective evaluation results of the in-vehicle noise based on the deep learning network model according to the validation results, or optimize the frequency band division of the in-vehicle noise spectrum data to obtain a training set again, and train the deep learning network model again.

[0041] In a third aspect, an in-vehicle noise evaluation and prediction implementation device is provided. The in-vehicle noise evaluation and prediction implementation device can be a personal computer (PC), a notebook computer, a server, or any other device having a data processing function.

[0042] In an embodiment, refer to Figure 3 , Figure 3 FIG. 2 is a schematic diagram of a hardware structure of an in-vehicle noise evaluation and prediction implementation device according to an embodiment of the present application. As shown in FIG. 2, the in-vehicle noise evaluation and prediction implementation device can include a processor, a memory, a communication interface, and a communication bus.

[0043] The communication bus can be of any type, and is configured to interconnect the processor, the memory, and the communication interface.

[0044] The communication interface includes an input / output (I / O) interface, a physical interface, and a logical interface, and is configured to interconnect devices within the in-vehicle noise evaluation and prediction implementation device, and to interconnect the in-vehicle noise evaluation and prediction implementation device with other devices (e.g., other computing devices or user devices). The physical interface can be an Ethernet interface, a fiber interface, an ATM interface, or the like. The user device can be a display (Display), a keyboard (Keyboard), or the like.

[0045] ​The memory can be various types of storage media, such as random access memory (RAM), read-only memory (ROM), non-volatile RAM (NVRAM), flash memory, optical storage, hard disk, programmable ROM (PROM), erasable PROM (EPROM), electrically erasable PROM (EEPROM), and the like.

[0046] The processor can be a general-purpose processor, which can invoke the in-vehicle noise evaluation prediction implementation program stored in the memory and execute the in-vehicle noise evaluation prediction implementation method provided by the embodiments of the present application. For example, the general-purpose processor can be a central processing unit (CPU). The method executed when the in-vehicle noise evaluation prediction implementation program is invoked can refer to each embodiment of the in-vehicle noise evaluation prediction implementation method of the present application, which will not be described here.

[0047] Those skilled in the art can understand that Figure 3 The hardware structure shown in the above-mentioned figures does not constitute a limitation on the present application, and can include more or fewer components than those shown, or combine certain components, or different arrangement of components.

[0048] In a fourth aspect, the embodiments of the present application further provide a computer readable storage medium.

[0049] The computer readable storage medium of the present application stores the in-vehicle noise evaluation prediction implementation program therein, wherein the in-vehicle noise evaluation prediction implementation program, when executed by the processor, implements the steps of the in-vehicle noise evaluation prediction implementation method as described above.

[0050] The method implemented when the in-vehicle noise evaluation prediction implementation program is executed can refer to each embodiment of the in-vehicle noise evaluation prediction implementation method of the present application, which will not be described here.

[0051] The terms “include,” “comprise,” “have,” and any variations thereof, in the Specification and in the Claims of the present application, and the above-mentioned drawings, are intended to cover a non-exclusive inclusion. For example, a process, method, system, product, or device that includes a list of steps or units is not limited to the listed steps or units, but can optionally further include steps or units not listed, or can optionally further include other steps or units inherent to such processes, methods, products, or devices. The terms “first,” “second,” and “third” and the like descriptions are used to distinguish different objects, and do not represent a sequence or limit the types of “first,” “second,” and “third.”

[0052] In the description of the embodiments of the present application, “exemplary”, “for example”, or “for instance” is used to represent an example, an illustration, or a description. Any embodiment or design scheme described as “exemplary”, “for example”, or “for instance” in the embodiments of the present application should not be interpreted as more preferred or more advantageous than other embodiments or design schemes. Rather, the words “exemplary”, “for example”, or “for instance” are intended to present the relevant concept in a specific manner.

[0053] In the description of the embodiments of the present application, unless otherwise specified, “ / ” represents the meaning of or, for example, A / B can represent A or B; “and / or” in the text only represents a description of the relationship between the associated objects, which means that there can be three relationships, for example, A and / or B can represent the three cases of A alone, A and B together, and B alone. In addition, in the description of the embodiments of the present application, “multiple” means two or more than two.

[0054] In some of the processes described in the embodiments of the present application, a plurality of operations or steps are included in a specific order, but it should be understood that these operations or steps can be executed or performed in parallel or in an order different from that in which they appear in the embodiments of the present application. The serial number of the operation is only used to distinguish different operations, and the serial number itself does not represent any execution order. In addition, these processes can include more or fewer operations, and these operations or steps can be executed in sequence or in parallel, and these operations or steps can be combined.

[0055] From the above description of the embodiments, those skilled in the art can clearly understand that the above-mentioned embodiment method can be realized by means of software and a general hardware platform as required, of course, it can also be realized by hardware, but in many cases the former is a better embodiment. Based on such understanding, the technical solutions of the present application can be embodied in the form of a software product, which is stored in a storage medium (such as a ROM / RAM, a magnetic disk, an optical disk) as described above, and includes a plurality of instructions for causing a terminal device to execute the methods described in the embodiments of the present application.

[0056] The preferred embodiments of the present application have been described above with the illustrated embodiments, and are not intended to limit the scope of patent protection for the present application. Any equivalent structure or equivalent process variations, which directly or indirectly incorporate the contents of the specification and drawings of the present application, are also intended to be included within the scope of patent protection for the present application.

Claims

1. A method for evaluating and predicting in-vehicle noise, characterized in that: The method for realizing the evaluation and prediction of vehicle interior noise includes: Conducting interior noise testing on the target vehicle to obtain interior noise spectrum data, and obtaining subjective evaluation results based on a subjective evaluation method; The in-car noise spectrum data is divided into frequency bands and labeled using the subjective evaluation results of the corresponding in-car noise spectrum data to obtain a dataset for the construction of training and validation sets. The training set is used to train the constructed deep learning network model, and the validation set is used to validate the trained deep learning network model; According to the verification results, the subjective evaluation results of the in-vehicle noise are predicted based on the deep learning network model, or the frequency band division of the in-vehicle noise spectrum data is optimized to obtain a training set again, and the deep learning network model is trained again.

2. The method for realizing in-vehicle noise evaluation prediction according to claim 1, characterized in that: The in-vehicle noise test of the target vehicle is performed to obtain in-vehicle noise spectrum data, and a subjective evaluation result is obtained based on a subjective evaluation method, specifically including: Determine the test frequency range of the in-vehicle noise test, perform an in-vehicle noise test on the target vehicle to obtain multiple noise test data, and perform spectrum analysis on each individual noise test data to obtain in-vehicle noise spectrum data; While performing the in-car noise test to obtain noise test data, a subjective evaluation of the in-car noise is performed to obtain a subjective evaluation result, and the subjective evaluation result corresponds one-to-one with the in-car noise spectrum data.

3. The method for realizing in-vehicle noise evaluation prediction according to claim 1, characterized in that: The method of predicting the subjective evaluation results of the interior noise of the vehicle based on the deep learning network model according to the verification results, or optimizing the frequency band division of the interior noise spectrum data to obtain a training set again, and retraining the deep learning network model, specifically includes: RMES value and R2 value of the verification results of the deep learning network model trained according to the verification set: If the RMES value is greater than the set value or the R2 value is less than the preset value, the frequency band division of the interior noise spectrum data is optimized, and a data set is obtained again to train the deep learning network model. Then, based on the retrained deep learning network model, the subjective evaluation results of the interior noise are predicted; If the RMES value is not greater than the set value and the R2 value is not less than the preset value, the deep learning network model is used to predict the subjective evaluation results of the interior noise.

4. The method for realizing in-vehicle noise evaluation prediction according to claim 1, wherein: The process of dividing the in-car noise spectrum data into frequency bands and labeling the data using the subjective evaluation results of the corresponding in-car noise spectrum data to obtain a data set for constructing a training set and a validation set specifically includes: Divide the vehicle interior noise spectrum data into frequency bands based on a 1 / 3 octave division method to obtain the frequency bands of each vehicle interior noise spectrum data; Based on the amplitude of each frequency band of the current in-car noise spectrum data, a 32-dimensional vector is constructed as a test data, and the label of the current test data is the subjective evaluation result corresponding to the current in-car noise spectrum data. Multiple test data are obtained to obtain a data set; The data set is divided according to the preset ratio to construct the training set and validation set.

5. The method for realizing in-vehicle noise evaluation prediction according to claim 4, characterized in that: The optimization of the frequency band division of the vehicle interior noise spectrum data is as follows: For the 32 frequency bands divided into 1 / 3 octaves, a preset number of frequency bands are randomly selected each time, and the fl value of each selected frequency band is updated based on the first preset algorithm, and the fu value of each selected frequency band is updated based on the first set algorithm or the second set algorithm to obtain a new frequency band division method, and finally obtain multiple new frequency band division methods; Based on the obtained new frequency band division methods, all vehicle interior noise spectrum data are divided into frequency bands and input into the trained deep learning network model to obtain prediction results. The new frequency band division method with the highest accuracy is selected as the final frequency band division method. The final frequency band division method is used to divide the vehicle interior noise spectrum data into frequency bands to obtain the data set again; If the selected frequency band includes the first frequency band among the 32 frequency bands, the fl value of the first frequency band is not updated; if the selected frequency band includes the last frequency band among the 32 frequency bands, the fu value of the last frequency band is not updated.

6. The method for realizing vehicle interior noise evaluation and prediction according to claim 5, characterized in that: The first preset algorithm is fla(i)=fu(i-1), where fla(i) represents the updated fl value of the i-th frequency band, and fu(i-1) represents the fu value of the i-1th frequency band; The first setting algorithm is fua(i)=fu(i)-(fu(i)-fl(i))×0.1, where fua(i) represents the updated fu value of the i-th frequency band, fu(i) represents the fu value of the i-th frequency band, and fl(i) represents the fl value of the i-th frequency band; The second setting algorithm is fua(i)=fu(i)+(fu(i)-fl(i))×0.

1.

7. The method for realizing vehicle interior noise evaluation and prediction according to claim 1, wherein: The deep learning network model is DNN, and the nonlinear activation function of DNN is Relu; When training the deep learning network model, the optimizer selects the stochastic gradient descent method and the loss function selects L2-Loss.

8. A device for realizing vehicle interior noise evaluation and prediction, characterized in that: The vehicle interior noise evaluation prediction implementation device comprises: An acquisition module is used to perform an interior noise test on a target vehicle to obtain interior noise spectrum data and obtain a subjective evaluation result based on a subjective evaluation method; A construction module is used to divide the in-vehicle noise spectrum data into frequency bands and label the data using the subjective evaluation results of the corresponding in-vehicle noise spectrum data to obtain a dataset for the construction of training and validation sets; A training module, which is used to train the constructed deep learning network model using the training set and to verify the trained deep learning network model using the verification set; An execution module is used to predict the subjective evaluation results of the vehicle interior noise based on the deep learning network model according to the verification results, or to optimize the frequency band division of the vehicle interior noise spectrum data to obtain a training set again and train the deep learning network model again.

9. A device for realizing vehicle interior noise evaluation and prediction, characterized in that: The in-vehicle noise evaluation prediction implementation device includes a processor, a memory, and an in-vehicle noise evaluation prediction implementation program stored on the memory and executable by the processor. When the in-vehicle noise evaluation prediction implementation program is executed by the processor, the steps of the in-vehicle noise evaluation prediction implementation method as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a vehicle interior noise evaluation prediction implementation program, wherein when the vehicle interior noise evaluation prediction implementation program is executed by the processor, the steps of the vehicle interior noise evaluation prediction implementation method according to any one of claims 1 to 7 are implemented.

Images