Engine sound quality evaluation method, system, equipment and medium

Abstract

The invention provides an engine sound quality evaluation method, system and device and a medium, and belongs to the technical field of engines. The method comprises the steps that firstly, multichannel noise signals of an engine under multiple working conditions are collected, acoustic parameters such as loudness, sharpness, roughness and fluctuation of the noise signals are calculated, and comprehensive acoustic parameter vectors of all the working conditions are obtained after synthesis; meanwhile, reference scores of the corresponding noise signals are obtained based on an expert audition test and synthesized into a comprehensive reference score; and then, taking the comprehensive acoustic parameter vector as input, taking the comprehensive reference score as a target, and training a neural network to construct a sound quality evaluation model. And finally, for a to-be-evaluated working condition, a score can be output by the model only by inputting the comprehensive acoustic parameter vector of the to-be-evaluated working condition. According to the method, the neural network model based on objective acoustic parameters and standardized subjective scoring is constructed, so that the beneficial effect of efficiently, stably, automatically and objectively evaluating the sound quality of the engine, which is highly consistent with human auditory perception, is achieved.

Description

Technical Field

[0001] This invention belongs to the field of engine technology, and more specifically relates to a method, system, equipment and medium for evaluating engine sound quality. Background Technology

[0002] In the field of engine design and manufacturing, with continuous technological advancements and increasingly demanding user expectations for driving experiences, the acoustic characteristics generated during engine operation—namely, sound quality—have become a key indicator for evaluating overall product performance and competitiveness. Sound quality transcends the traditional noise control focus solely on "sound pressure level," placing greater emphasis on the subjective experience of sound, such as whether it is pleasant, rough, or sharp. Therefore, developing a scientific and effective method for evaluating sound quality is of great significance for optimizing engine NVH (noise, vibration, and harshness) performance and meeting market demands.

[0003] Currently, mainstream sound quality evaluation methods in the industry focus on the measurement and analysis of objective physical indicators. This method typically relies on specialized noise testing equipment and signal analysis software to collect engine noise signals and calculate a series of psychoacoustic parameters such as loudness, sharpness, roughness, and fluctuation to quantify acoustic characteristics. While this purely objective analysis method has the advantages of stable results and high repeatability, its fundamental limitation lies in equating sound quality—an attribute inherently closely related to human auditory perception—completely with physical quantities that can be directly measured by instruments. However, the "quality" of sound is ultimately determined by human hearing and psychological perception. The same combination of objective parameters may produce drastically different subjective evaluations due to individual differences, cultural backgrounds, or usage scenarios. Therefore, relying solely on instrument data cannot accurately reflect the true user experience of sound, potentially leading to a disconnect between evaluation results and actual human perception.

[0004] On the other hand, while directly adopting users' subjective evaluations can reflect auditory perception at its source, traditional subjective evaluation methods have significant drawbacks. Typically, this method involves organizing reviewers to listen to and score sound samples. However, due to differences in reviewers' auditory sensitivity, personal preferences, experience, and current physical and mental state, the evaluation results often exhibit strong subjectivity, arbitrariness, and dispersion. This "poor consistency" problem makes subjective evaluation data difficult to use directly for precise engineering development and comparative analysis; its reliability and stability cannot meet the stringent requirements for quantifiable and comparable evaluation results in product development.

[0005] In summary, existing engine sound quality evaluation technologies suffer from a dilemma: objective measurement is divorced from subjective perception, while subjective evaluation lacks consistency with objectivity. These two aspects are disconnected, and neither can provide a comprehensive evaluation solution that is both scientifically reliable and accurately reflects the user's auditory experience. Summary of the Invention

[0006] To address the above problems, the present invention aims to provide a method, system, device, and medium for evaluating engine sound quality. By constructing a neural network mapping model based on objective acoustic parameters and standardized subjective scores, an efficient, stable, and highly consistent automated objective evaluation of engine sound quality with human auditory perception is achieved.

[0007] To achieve the above objectives, the present invention employs the following technical solution: In a first aspect, embodiments of this application provide a method for evaluating engine sound quality, including: Collect multi-channel noise signal data from multiple engines under various preset operating conditions to form a noise signal sample set; For the noise signal data under each preset operating condition in the sample set, acquire or calculate its acoustic parameters, including loudness. Sharpness Roughness and volatility ; For each preset operating condition in the sample set, the acoustic parameters of all channels under that operating condition are synthesized using a data synthesis algorithm, and the synthesized loudness is calculated. Synthetic sharpness Synthetic roughness and synthetic volatility Generate a comprehensive acoustic parameter vector representing each preset working condition; Based on an expert evaluation mechanism, a reference score value is obtained corresponding to the noise signal data under each preset working condition in the noise signal sample set. Furthermore, the reference score values ​​of all channels under the same working condition are synthesized using a data synthesis algorithm to obtain the comprehensive reference score for each working condition. ; Using the comprehensive acoustic parameter vector as the input sample, and the corresponding comprehensive reference score To generate a sound quality evaluation model, train a neural network model to achieve the target output. For the target operating condition of the engine to be evaluated, its comprehensive acoustic parameter vector is obtained and input into the sound quality evaluation model, and the model outputs a sound quality score. .

[0008] In an optional implementation, the step of collecting multi-channel noise signal data from multiple engines under various preset operating conditions to form a noise signal sample set includes: In the NVH (Noise, Vibration, and Harshness) silencing test chamber, a 9-point noise testing method was used to collect noise signal data for engines with different displacements from 1.2L to 13L under full-speed, full-throttle conditions, full-throttle external characteristic conditions, full-throttle acceleration conditions, DRB braking conditions, and idling conditions. The noise signal data included: sound signal frequency x, modulation frequency... Modulation depth .

[0009] In one optional implementation, for the noise signal data under each preset operating condition in the sample set, acoustic parameters are acquired or calculated, including loudness. Sharpness Roughness and volatility ,include: For the noise signal data under each preset working condition in the sample set, the loudness N corresponding to the noise signal data is obtained through the Moore model; For the noise signal data under each preset working condition in the sample set, the corresponding sharpness is calculated using the following formula. Roughness and volatility :

[0010]

[0011]

[0012] in, The frequency of the sound signal; Bark is the loudness scale value. For the weighting function of the Bark domain, The modulation frequency of the noise signal. Let x be the modulation depth at frequency x.

[0013] In one optional implementation, the data synthesis algorithm includes: Regarding loudness Sharpness Roughness volatility and reference score The composite value of any parameter P in the formula is calculated using the following formula. :

[0014] in, This represents the total number of measurement point channels. For the first Parameters obtained from each measuring point channel The value of .

[0015] In an optional implementation, the reference score value obtained based on the expert evaluation mechanism corresponds to the noise signal data under each preset operating condition in the noise signal sample set. ,include: Obtained through hearing tests Each evaluator assigns a score to the same noise signal data in the training noise signal sample set, and a reference score is calculated using the following formula. :

[0016] in, The number of valid ratings and , For the first The scores given by each evaluator Index for evaluators; For all The highest score in the, For all The lowest score in the test.

[0017] In an optional implementation, the step of using a comprehensive acoustic parameter vector as input samples and corresponding comprehensive reference scores... To achieve the target output, a neural network model is trained to generate a sound quality evaluation model, including: Construct a backpropagation (BP) neural network comprising an input layer, at least one hidden layer, and an output layer; wherein the number of nodes in the input layer corresponds to the dimension of the synthesized acoustic parameter vector, and is used to receive the vector. The output layer consists of a single node, used to output a predicted score for sound quality. The neural network uses weight coefficients determined through training. and bias coefficient This makes its output Establish the following mapping relationship between the input and the input:

[0018] The comprehensive acoustic parameter vector and its corresponding comprehensive reference score are used. A training dataset is constructed and input into the neural network for iterative training; in each iteration, forward propagation is performed to calculate the predicted score. Calculate the predicted score using the loss function. With corresponding comprehensive reference score Mean square error between And the weight coefficients and bias coefficients in the neural network are adjusted using the backpropagation algorithm. When the average mean square error When the number of training iterations falls below a preset threshold or reaches a preset upper limit, training is terminated, the neural network parameters at this point are saved, and a sound quality evaluation model is generated.

[0019] In an optional implementation, the loss function includes:

[0020] in, The number of training samples in the training dataset. Let i be the predicted score for the i-th training sample. This is the comprehensive reference score for the i-th training sample.

[0021] Secondly, embodiments of this application also provide an engine sound quality evaluation system, including: The system includes: The sample data acquisition module is used to collect multi-channel noise signal data from multiple engines under various preset operating conditions, forming a noise signal sample set. The parameter calculation module is used to acquire or calculate the acoustic parameters of the noise signal data under each preset working condition in the sample set. The acoustic parameters include loudness. Sharpness Roughness and volatility ; The data synthesis module is used to synthesize the acoustic parameters of all channels under each preset working condition in the sample set using a data synthesis algorithm, and calculate the synthesized loudness. Synthetic sharpness Synthetic roughness and synthetic volatility Generate a comprehensive acoustic parameter vector representing each preset working condition; The reference score calculation module is used to obtain a reference score value corresponding to the noise signal data under each preset working condition in the noise signal sample set based on an expert evaluation mechanism. Furthermore, the reference score values ​​of all channels under the same working condition are synthesized using a data synthesis algorithm to obtain the comprehensive reference score for each working condition. ; The model training module is used to train samples with a comprehensive acoustic parameter vector as input and a corresponding comprehensive reference score. To generate a sound quality evaluation model, train a neural network model to achieve the target output. The real-time evaluation module is used to obtain the comprehensive acoustic parameter vector of the engine under the target operating condition to be evaluated, and input it into the sound quality evaluation model, and output a sound quality score through the model. .

[0022] Thirdly, embodiments of this application also provide an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the engine sound quality evaluation method as described in any of the above.

[0023] Fourthly, embodiments of this application also provide a storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the engine sound quality evaluation method as described in any of the above claims.

[0024] As can be seen from the above technical solutions, the present invention has the following advantages: The engine sound quality evaluation method provided in this application achieves automated, objective, repeatable, and efficient evaluation of engine sound quality by constructing and training a neural network model based on expert evaluation calibration and mapping of objective acoustic parameters. This method deeply correlates objective acoustic parameters collected from multiple measurement points and under multiple operating conditions with a comprehensive reference score obtained through a standardized process. The resulting evaluation model accurately simulates human auditory perception, effectively overcoming the inherent defects of existing technologies where objective evaluation is divorced from subjective experience, while purely subjective evaluation suffers from poor consistency and repeatability. Applying this method allows for rapid and stable acquisition of sound quality quantification results highly consistent with expert evaluations during product development and quality control stages, requiring only standardized objective data collection, significantly improving the engineering practicality and efficiency of the evaluation.

[0025] This application uses a comprehensive reference score obtained from expert listening tests as the training target of a neural network, enabling the final evaluation model to accurately encode the complex mapping relationship between human auditory perception and acoustic physical parameters. This fundamentally solves the drawback of traditional objective evaluation methods being detached from actual auditory experience, making the evaluation results possess both the rigor of objective data and the authenticity of subjective feelings.

[0026] This application transforms subjective opinions into stable data through a standardized scoring process and data synthesis algorithm, and trains a deterministic mathematical model. For the same objective input, the model always outputs the same score, completely solving the problem of evaluation results varying from person to person and from time to time, and providing a reliable and consistent quantitative benchmark for product performance comparison and quality control.

[0027] After model training is completed, this application only requires standardized objective data collection and calculation to evaluate new products or new operating conditions, and the evaluation results can be obtained instantly through the model. This realizes the transformation from manual evaluation to automatic prediction, significantly shortens the development cycle, reduces evaluation costs, and is more suitable for modern rapid iterative R&D processes.

[0028] This application ensures the accuracy and generalization ability of the evaluation model through a systematic data synthesis and model validation mechanism. Multi-point data synthesis technology effectively characterizes the spatial characteristics of the engine sound field, reducing random errors from single-point measurements. Simultaneously, rigorous training and validation processes, along with an MSE monitoring mechanism, ensure the model's prediction accuracy and stability on unknown data. This enables the model to not only fit the training data well but also make accurate judgments on new samples, demonstrating strong reliability for practical applications.

[0029] This application expands the standardized application scenarios and decision support value of acoustic quality evaluation. The generated generalized evaluation model can serve as a standard tool for horizontal comparison and vertical optimization of acoustic quality of different engine models under various operating conditions. Its output quantitative score can provide clear and intuitive optimization directions and goals for design improvements, thereby transforming the abstract concept of acoustic quality into concrete parameters that can drive engineering decisions, and strongly supporting the refinement and goal-oriented approach of product acoustic design. Attached Figure Description

[0030] To more clearly illustrate the technical solution of the present invention, the accompanying drawings used in the description will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0031] Figure 1 A flowchart illustrating the engine sound quality evaluation method provided in this application.

[0032] Figure 2 A schematic diagram illustrating the principle of the neural network model provided in this application.

[0033] Figure 3 This is a schematic diagram of the engine sound quality evaluation system provided in this application.

[0034] Figure 4 A schematic diagram of the structure of the electronic device provided in this application. Detailed Implementation

[0035] The various embodiments of this disclosure will be described more fully in the detailed steps of the engine sound quality evaluation method described below. This disclosure may have various embodiments, and adjustments and changes may be made therein. However, it should be understood that there is no intention to limit the various embodiments of this disclosure to the specific embodiments disclosed herein, but rather this disclosure should be understood to cover all adjustments, equivalents, and / or alternatives falling within the spirit and scope of the various embodiments of this disclosure.

[0036] In the following, the terms “comprising” or “may include”, which may be used in various embodiments of this disclosure, indicate the presence of the disclosed functions, operations, or elements, and do not limit the addition of one or more functions, operations, or elements. Furthermore, as used in various embodiments of this disclosure, the terms “comprising,” “having,” and their cognates are intended only to indicate a particular feature, number, step, operation, element, component, or combination of the foregoing, and should not be construed as primarily excluding the presence of one or more other features, numbers, steps, operations, elements, components, or combinations of the foregoing, or the possibility of adding one or more combinations of the foregoing.

[0037] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0038] Please see Figure 1 The diagram shows a flowchart of an engine sound quality evaluation method in a specific embodiment. The method includes: S1: Collect multi-channel noise signal data from multiple engines under various preset operating conditions to form a noise signal sample set.

[0039] In a specific implementation, in an NVH (Noise, Vibration, and Harshness) silencing test chamber, a 9-point noise testing method is used to collect noise signal data for engines with different displacements from 1.2L to 13L under full-speed, full-throttle conditions, full-throttle external characteristic conditions, full-throttle acceleration conditions, DRB braking conditions, and idling conditions. The noise signal data includes: sound signal frequency x, modulation frequency... Modulation depth .

[0040] The 9-point noise testing method uses 9 sound signal measurement points, located directly above the engine (9), at the front (free end), at the rear (flywheel end), on the left (facing the same direction as the free end, away from the flywheel end), on the right, at the upper right front, at the upper left front, at the upper right rear, and at the upper left rear. Each measurement point has one high-precision IEPE microphone, which is calibrated using a GRAS 42AA acoustic calibrator before testing.

[0041] S2: For the noise signal data under each preset working condition in the sample set, acquire or calculate its acoustic parameters, including loudness. Sharpness Roughness and volatility

[0042] In a specific implementation, for the noise signal data under each preset working condition in the sample set, the loudness N corresponding to the noise signal data is obtained through the Moore model; the unit is Son, which can be directly obtained through testing software according to the ANSI S3.4 standard.

[0043] For the noise signal data under each preset operating condition in the sample set, the sharpness is calculated using the Zwicker general calculation model and the following formula. :

[0044] in, Sound signal frequency, Bark is the loudness scale value (0-24). The weighting function for the Bark domain is defined as:

[0045] For the noise signal data under each preset working condition in the sample set, the corresponding roughness is calculated using the following formula. :

[0046] in The modulation frequency is the number of sound pressure fluctuations per unit time. The modulation depth at frequency x is the degree of change in sound pressure amplitude.

[0047] For the noise signal data under each preset operating condition in the sample set, the corresponding fluctuation is calculated using the following formula. :

[0048] Among them, volatility The unit is vacuum.

[0049] S3: For each preset operating condition in the sample set, the acoustic parameters of all channels under that operating condition are synthesized using a data synthesis algorithm, and the synthesized loudness is calculated. Synthetic sharpness Synthetic roughness and synthetic volatility This generates a comprehensive acoustic parameter vector representing each preset operating condition.

[0050] In a specific implementation, this step and subsequent steps all use a data synthesis algorithm to calculate the acoustic parameters or the synthesized reference score value.

[0051] Specifically, under a certain operating condition of the engine, because a 9-point method is used for testing, there are 9 different sound signals in a single operating condition. The following formula can be used to fit these sound signals. This formula is obtained by correcting the national standard nine-point sound pressure level calculation method. The synthesized loudness can be obtained through this formula. Synthetic sharpness Synthetic roughness Synthetic volatility Comprehensive reference score :

[0052] in, For the first Parameters obtained from each measuring point channel The value of . In practical applications, represent Or refer to the rating value .

[0053] S4: Based on an expert evaluation mechanism, obtain reference score values ​​corresponding to the noise signal data under each preset working condition in the noise signal sample set. Furthermore, the reference score values ​​of all channels under the same working condition are synthesized using a data synthesis algorithm to obtain the comprehensive reference score for each working condition. .

[0054] In a specific implementation, the first step is to obtain the information through a listening test. Each evaluator assigns a score to the same noise signal data in the training noise signal sample set. For example, 50 volunteers (25 men and 25 women) with over 5 years of driving experience and over 50,000 kilometers of driving mileage can be recruited to subjectively evaluate the noise signal data under each preset working condition in the noise signal sample set, obtaining corresponding reference scores. This method addresses, to some extent, the issues of strong subjectivity and large deviations in individual evaluation results in subjective assessments. The subjective evaluation uses a scale of 1-10, with 10 being the best and 1 the worst. The listening test scoring reference table is shown in Table 1 below: Table 1: Reference Table for Listening Test Scoring

[0055] Then, based on the obtained rating values, the reference rating value is calculated using the following formula. :

[0056] in, The number of valid ratings and , For the first The scores given by each evaluator Index for evaluators; For all The highest score in the, For all The lowest score in the test.

[0057] Finally, the reference score values ​​of all channels under the same working condition are synthesized using the data synthesis algorithm disclosed in step S4 to obtain the comprehensive reference score for each working condition. .

[0058] S5: Using the comprehensive acoustic parameter vector as input sample and the corresponding comprehensive reference score. To achieve the target output, a neural network model is trained to generate a sound quality evaluation model.

[0059] In a specific implementation, a backpropagation (BP) neural network is constructed, comprising an input layer, at least one hidden layer, and an output layer; wherein the number of nodes in the input layer corresponds to the dimension of the synthesized acoustic parameter vector, and is used to receive the vector. The output layer consists of a single node, used to output a predicted score for sound quality. The neural network uses weight coefficients determined through training. and bias coefficient This makes its output Establish the following mapping relationship between the input and the input:

[0060] For example, the test involved 10 engines, and data was collected from each engine under full-speed, full-throttle conditions, full-throttle external characteristic conditions, full-throttle acceleration conditions, DRB braking conditions, and idling conditions, totaling 50 sets of data. Of these, 30 sets were used to train the model, and 20 sets were used for validation. A sound quality scoring model Y, based on both subjective and objective factors, was derived. This model combines the advantages of both subjective and objective approaches, representing both the merits and demerits of objective indicators and reflecting the user's actual experience. A simplified diagram of the neural network-based model building method is shown below. Figure 2 As shown.

[0061] in , , , These are weighting coefficients. , where are the hidden layer activation functions and Y is the output function.

[0062] Then, the integrated acoustic parameter vector and its corresponding integrated reference score are... A training dataset is constructed and input into the neural network for iterative training; in each iteration, forward propagation is performed to calculate the predicted score. Calculate the predicted score using the loss function. With corresponding comprehensive reference score Mean square error between The weight coefficients and bias coefficients in the neural network are adjusted using the backpropagation algorithm.

[0063] In this study, an evaluation model Y was established using 30 sets of data through a neural network. The reliability of the model was then verified using the Mean Mean Square Error (MSE) function. A smaller MSE value indicates a higher correlation between the model's simulated output and the overall reference score. The closer they are, the more the loss function includes:

[0064] in, The number of training samples in the training dataset. Let i be the predicted score for the i-th training sample. This is the comprehensive reference score for the i-th training sample.

[0065] When the average mean square error When the number of training iterations falls below a preset threshold or reaches a preset upper limit, training is terminated, the neural network parameters at this point are saved, and a sound quality evaluation model is generated.

[0066] S6: For the target operating condition of the engine to be evaluated, obtain its comprehensive acoustic parameter vector and input it into the sound quality evaluation model, and output a sound quality score through the model. .

[0067] In a specific implementation, a noise signal sample set from the target operating condition of the engine to be evaluated is obtained. The corresponding comprehensive acoustic parameter vector is generated using the processing method described in the preceding steps of this method, and then input into the sound quality evaluation model. The model outputs a sound quality score. Only the synthesized loudness needs to be measured. Synthetic sharpness Synthetic roughness Synthetic volatility By inputting the data into model Y, a sound quality score can be obtained, which should be basically consistent with actual user evaluations.

[0068] In this embodiment, a neural network model is constructed with multi-condition integrated acoustic parameters as input and standardized expert scores as training targets. This model quantitatively integrates objective acoustic measurements with subjective auditory perception, thereby successfully overcoming the core defects of traditional evaluation methods, such as poor consistency of subjective evaluation and objective evaluation being detached from actual experience. Ultimately, this achieves an efficient, stable, repeatable, and highly consistent automated objective evaluation of engine sound quality with human hearing, significantly improving the engineering practicality and product development guidance value of the evaluation work.

[0069] like Figure 3 As shown, the following are embodiments of the engine sound quality evaluation system provided in this disclosure. This system and the engine sound quality evaluation methods in the above embodiments belong to the same inventive concept. For details not described in detail in the embodiments of the engine sound quality evaluation system, please refer to the embodiments of the above engine sound quality evaluation methods.

[0070] An engine sound quality evaluation system, comprising: The sample data acquisition module is used to collect multi-channel noise signal data from multiple engines under various preset operating conditions, forming a noise signal sample set.

[0071] The parameter calculation module is used to acquire or calculate the acoustic parameters of the noise signal data under each preset working condition in the sample set. The acoustic parameters include loudness. Sharpness Roughness and volatility

[0072] The data synthesis module is used to synthesize the acoustic parameters of all channels under each preset working condition in the sample set using a data synthesis algorithm, and calculate the synthesized loudness. Synthetic sharpness Synthetic roughness and synthetic volatility This generates a comprehensive acoustic parameter vector representing each preset operating condition.

[0073] The reference score calculation module is used to obtain a reference score value corresponding to the noise signal data under each preset working condition in the noise signal sample set based on an expert evaluation mechanism. Furthermore, the reference score values ​​of all channels under the same working condition are synthesized using a data synthesis algorithm to obtain the comprehensive reference score for each working condition. .

[0074] The model training module is used to train samples with a comprehensive acoustic parameter vector as input and a corresponding comprehensive reference score. To achieve the target output, a neural network model is trained to generate a sound quality evaluation model.

[0075] The real-time evaluation module is used to obtain the comprehensive acoustic parameter vector of the engine under the target operating condition to be evaluated, and input it into the sound quality evaluation model, and output a sound quality score through the model. .

[0076] The engine sound quality evaluation system provided in this embodiment achieves automated and quantitative evaluation of engine sound quality by constructing a neural network model based on the fusion training of objective acoustic parameters from multiple operating conditions and multiple measurement points with standardized subjective scoring data. This system effectively overcomes the inherent defects of traditional purely objective evaluation, which is detached from the real perception of human ears, and purely subjective evaluation, which suffers from poor consistency and repeatability. The final generated model can stably and efficiently output quantitative scores that are highly consistent with the auditory perception of experts, significantly improving the engineering practical efficiency and reliability of the evaluation work.

[0077] Figure 4 A schematic diagram of the hardware structure of an electronic device for implementing various embodiments of the present invention.

[0078] The engine sound quality evaluation method provided in this application embodiment can be applied to electronic devices. Those skilled in the art will understand that the electronic device structure involved in the embodiments of this invention does not constitute a limitation on the electronic device. An electronic device may include more or fewer components than illustrated, or combine certain components, or have different component arrangements. In the embodiments of this invention, the electronic device includes, but is not limited to, laptop computers, desktop computers, workbenches, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the embodiments of this application described and / or claimed herein.

[0079] Electronic devices may include processors, external memory interfaces, internal memory, universal serial bus (USB) interfaces, charging management modules, power management modules, batteries, wireless communication modules, audio modules, speakers, microphones, sensor modules, buttons, cameras, displays, and SIM card interfaces, etc.

[0080] A processor may include one or more processing units, such as: a central processing unit (CPU), an application processor (AP), a modem processor, a graphics processing unit (GPU), an image signal processor (ISP), a controller, memory, a video codec, a digital signal processor (DSP), a baseband processor, and / or a neural network processing unit (NPU). Different processing units may be independent devices or integrated into one or more processors.

[0081] The processor can serve as the nerve center and command center of an electronic device. The controller can generate operation control signals based on the instruction opcode and timing signals to control the fetching and execution of instructions.

[0082] The processor may also include memory for storing instructions and data. In some embodiments, the memory in the processor is a cache memory. This memory can store instructions or data that the processor has just used or that are used repeatedly. If the processor needs to use the instruction or data again, it can retrieve it directly from this memory. This avoids repeated accesses, reduces processor latency, and thus improves system efficiency.

[0083] An external storage interface (ESI) can be used to connect external memory cards, such as microSD cards, to expand the storage capacity of electronic devices. The external memory card communicates with the processor through the ESI to perform data storage functions, such as saving music and video files on the external memory card.

[0084] Internal memory can be used to store computer executable program code, which includes instructions. The processor executes various functional applications and data processing of electronic devices by running the instructions stored in internal memory. Internal memory can include a program storage area and a data storage area. Internal memory can include high-speed random access memory, and can also include non-volatile memory, such as at least one disk storage device, flash memory device, universal flash storage (UFS), etc.

[0085] Wireless communication functionality in electronic devices can be achieved through antennas, wireless communication modules, modem processors, and baseband processors.

[0086] Wireless communication modules can provide solutions for wireless communication applications in electronic devices, including wireless local area networks (WLANs) (such as wireless fidelity (Wi-Fi) networks), Bluetooth (BT), global navigation satellite system (GNSS), frequency modulation (FM), near field communication (NFC), and infrared (IR) technologies.

[0087] Electronic devices can implement audio functions through audio modules, speakers, receivers, microphones, headphone jacks, and application processors.

[0088] Electronic devices can achieve shooting functions through ISPs, cameras, video codecs, GPUs, displays, and application processors.

[0089] Electronic devices can achieve display functions through GPUs, displays, and application processors.

[0090] A GPU is a microprocessor for image processing, connected to the display screen and application processor. GPUs are used to perform mathematical and geometric calculations for graphics rendering. A processor may include one or more GPUs, which execute program instructions to generate or modify display information.

[0091] A display screen is used to display images, videos, etc. A display screen includes a display panel.

[0092] The aforementioned electronic device realizes the beneficial effect of the engine sound quality evaluation method of this application by constructing a neural network model that integrates objective acoustic parameters under multiple operating conditions and standardized subjective scores, thereby objectifying and quantifying subjective auditory perception, and thus achieving efficient, stable and highly consistent automated evaluation of engine sound quality with human perception.

[0093] The storage medium provided in this application stores a program product capable of implementing an engine sound quality evaluation method.

[0094] Engine sound quality evaluation methods include: Collect multi-channel noise signal data from multiple engines under various preset operating conditions to form a noise signal sample set; For the noise signal data under each preset operating condition in the sample set, acquire or calculate its acoustic parameters, including loudness. Sharpness Roughness and volatility ; For each preset operating condition in the sample set, the acoustic parameters of all channels under that operating condition are synthesized using a data synthesis algorithm, and the synthesized loudness is calculated. Synthetic sharpness Synthetic roughness and synthetic volatility Generate a comprehensive acoustic parameter vector representing each preset working condition; Based on an expert evaluation mechanism, a reference score value is obtained corresponding to the noise signal data under each preset working condition in the noise signal sample set. Furthermore, the reference score values ​​of all channels under the same working condition are synthesized using a data synthesis algorithm to obtain the comprehensive reference score for each working condition. ; Using the comprehensive acoustic parameter vector as the input sample, and the corresponding comprehensive reference score To generate a sound quality evaluation model, train a neural network model to achieve the target output. For the target operating condition of the engine to be evaluated, its comprehensive acoustic parameter vector is obtained and input into the sound quality evaluation model, and the model outputs a sound quality score. .

[0095] In some possible implementations, the engine sound quality evaluation method of this disclosure can be implemented as a program product comprising program code that, when run on a terminal device, causes the terminal device to perform the steps described in the "Exemplary Methods" section of this specification according to various exemplary embodiments of this disclosure.

[0096] The storage medium disclosed herein may be any combination of one or more readable media. A readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples (a non-exhaustive list) of readable storage media include: an electrical connection having one or more wires, a portable disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof.

[0097] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for evaluating engine sound quality, characterized in that, include: Collect multi-channel noise signal data from multiple engines under various preset operating conditions to form a noise signal sample set; For the noise signal data under each preset operating condition in the sample set, acquire or calculate its acoustic parameters, including loudness. Sharpness Roughness and volatility ; For each preset operating condition in the sample set, the acoustic parameters of all channels under that operating condition are synthesized using a data synthesis algorithm, and the synthesized loudness is calculated. Synthetic sharpness Synthetic roughness and synthetic volatility Generate a comprehensive acoustic parameter vector representing each preset working condition; Based on an expert evaluation mechanism, a reference score value is obtained corresponding to the noise signal data under each preset working condition in the noise signal sample set. Furthermore, the reference score values ​​of all channels under the same working condition are synthesized using a data synthesis algorithm to obtain the comprehensive reference score for each working condition. ; Using the comprehensive acoustic parameter vector as the input sample, and the corresponding comprehensive reference score To generate a sound quality evaluation model, train a neural network model to achieve the target output. For the target operating condition of the engine to be evaluated, its comprehensive acoustic parameter vector is obtained and input into the sound quality evaluation model, and the model outputs a sound quality score. .

2. The engine sound quality evaluation method according to claim 1, characterized in that, The acquisition of multi-channel noise signal data from multiple engines under various preset operating conditions constitutes a noise signal sample set, including: In the NVH (Noise, Vibration, and Harshness) silencing test chamber, a 9-point noise testing method was used to collect noise signal data for engines with different displacements from 1.2L to 13L under full-speed, full-throttle conditions, full-throttle external characteristic conditions, full-throttle acceleration conditions, DRB braking conditions, and idling conditions. The noise signal data included: sound signal frequency x, modulation frequency... Modulation depth .

3. The engine sound quality evaluation method according to claim 2, characterized in that, For the noise signal data under each preset working condition in the sample set, acoustic parameters are acquired or calculated, including loudness. Sharpness Roughness and volatility ,include: For the noise signal data under each preset working condition in the sample set, the loudness N corresponding to the noise signal data is obtained through the Moore model; For the noise signal data under each preset working condition in the sample set, the corresponding sharpness is calculated using the following formula. Roughness and volatility : in, The frequency of the sound signal; Bark is the loudness scale value. For the weighting function of the Bark domain, The modulation frequency of the noise signal. Let x be the modulation depth at frequency x.

4. The engine sound quality evaluation method according to claim 1, characterized in that, The data synthesis algorithm includes: Regarding loudness Sharpness Roughness volatility and reference score The composite value of any parameter P in the formula is calculated using the following formula. : in, This represents the total number of measurement point channels. For the first Parameters obtained from each measuring point channel The value of .

5. The engine sound quality evaluation method according to claim 1, characterized in that, The reference score value corresponding to the noise signal data under each preset working condition in the noise signal sample set is obtained based on the expert evaluation mechanism. ,include: Obtained through hearing tests Each evaluator assigns a score to the same noise signal data in the training noise signal sample set, and a reference score is calculated using the following formula. : in, The number of valid ratings and , For the first The scores given by each evaluator Index for evaluators; For all The highest score in the competition. For all The lowest score in the test.

6. The engine sound quality evaluation method according to claim 1, characterized in that, The input sample is a comprehensive acoustic parameter vector, with the corresponding comprehensive reference score. To achieve the target output, a neural network model is trained to generate a sound quality evaluation model, including: Construct a backpropagation (BP) neural network comprising an input layer, at least one hidden layer, and an output layer; wherein the number of nodes in the input layer corresponds to the dimension of the synthesized acoustic parameter vector, and is used to receive the vector. The output layer consists of a single node, used to output a predicted score for sound quality. The neural network uses weight coefficients determined through training. and bias coefficient This makes its output Establish the following mapping relationship between the input and the input: The comprehensive acoustic parameter vector and its corresponding comprehensive reference score are used. A training dataset is constructed and input into the neural network for iterative training; in each iteration, forward propagation is performed to calculate the predicted score. Calculate the predicted score using the loss function. With corresponding comprehensive reference score Mean square error between And the weight coefficients and bias coefficients in the neural network are adjusted using the backpropagation algorithm. When the average mean square error When the number of training iterations falls below a preset threshold or reaches a preset upper limit, training is terminated, the neural network parameters at this point are saved, and a sound quality evaluation model is generated.

7. The engine sound quality evaluation method according to claim 6, characterized in that, The loss function includes: in, The number of training samples in the training dataset. Let i be the predicted score for the i-th training sample. This is the comprehensive reference score for the i-th training sample.

8. An engine sound quality evaluation system, characterized in that, The system employs the engine sound quality evaluation method as described in any one of claims 1 to 7; The system includes: The sample data acquisition module is used to collect multi-channel noise signal data from multiple engines under various preset operating conditions, forming a noise signal sample set. The parameter calculation module is used to acquire or calculate the acoustic parameters of the noise signal data under each preset working condition in the sample set. The acoustic parameters include loudness. Sharpness Roughness and volatility ; The data synthesis module is used to synthesize the acoustic parameters of all channels under each preset working condition in the sample set using a data synthesis algorithm, and calculate the synthesized loudness. Synthetic sharpness Synthetic roughness and synthetic volatility Generate a comprehensive acoustic parameter vector representing each preset working condition; The reference score calculation module is used to obtain a reference score value corresponding to the noise signal data under each preset working condition in the noise signal sample set based on an expert evaluation mechanism. Furthermore, the reference score values ​​of all channels under the same working condition are synthesized using a data synthesis algorithm to obtain the comprehensive reference score for each working condition. ; The model training module is used to train samples with a comprehensive acoustic parameter vector as input and a corresponding comprehensive reference score. To generate a sound quality evaluation model, train a neural network model to achieve the target output. The real-time evaluation module is used to obtain the comprehensive acoustic parameter vector of the engine under the target operating condition to be evaluated, and input it into the sound quality evaluation model, and output a sound quality score through the model. .

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the engine sound quality evaluation method as described in any one of claims 1 to 7.

10. A storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the engine sound quality evaluation method as described in any one of claims 1 to 7.

Images