Multi-environment hearing aid gain prediction method and system, processing equipment and storage medium

Abstract

The invention relates to a multi-environment hearing aid gain prediction method and system, a processing device and a storage medium, and the method is characterized in that the method comprises the steps: obtaining an audiogram of a patient; inputting different hearing thresholds, different input sound pressure levels and environments in the audiogram of the patient into a trained multi-environment hearing aid gain prediction model to obtain prescription gain prediction results corresponding to the patient under different frequencies; the multi-environment hearing aid gain prediction model is constructed based on an XGBoost algorithm, a grid search method and a genetic algorithm, adaptive change of hearing aid gains in different environments can be realized, the prediction precision is high, the generalization performance is high, and the method can be widely applied to the field of gain compensation of hearing aid fitting.

Description

Technical Field

[0001] The present invention relates to the field of gain compensation for hearing aid fitting, and particularly to a multi-environment hearing aid gain prediction method, system, processing device, and storage medium. Background Art

[0002] Currently, 1.5 billion people worldwide are affected by hearing loss to varying degrees. There is a large number of hearing-impaired people in China, with a prevalence rate of 1.62%. With the intensification of population aging, the number of elderly hearing-impaired patients is increasing, and the demand for hearing aids is also increasing. A hearing aid is a rehabilitation aid to help hearing-impaired patients restore their hearing. Its core method is that the fitting specialist selects a suitable gain scheme for sounds of different loudness and frequencies according to the hearing condition of the hearing-impaired patient on the fitting software, compensates for the patient's hearing loss, and then helps the user hear and understand sounds better. Currently, the gain determination methods for hearing aid fitting include the NAL_NAL1, NAL-NAL2 formulas, the DSL-V5 formula, and the gain prediction of hearing aids based on machine learning. Among them, the NAL-NAL2 formula and the DSL-V5 formula are the main formulas currently used for hearing aid fitting, and the hearing aid fitting formula based on machine learning is also gradually emerging.

[0003] The NAL-NL2 formula is a non-linear hearing aid fitting prescription formula released by the National Acoustic Laboratories (NAL) of Australia. It is the main way to predict the gain of adult hearing aids. Its predecessor is the NAL-NL1. Their goals are both to maximize speech intelligibility. Compared with the NAL-NL1 formula, the NAL-NL2 formula gives a slightly higher compression ratio to patients with mild to moderate hearing loss, and takes into account the influence of factors such as the age and hearing aid wearing history of hearing-impaired patients on the gain applied by the fitting formula. The DSL-V5 formula is a hearing aid gain formula for predicting hearing-impaired patients as children. The purpose of this formula is to enable hearing aid wearers to obtain the maximum audibility in each frequency region, and particularly integrates testing methods for infants and young children, such as: auditory brainstem response (ABR), real-ear coupler difference (RECD), etc. It is a prescription formula recognized and widely applicable to infants, young children, and children in the industry; Machine learning is a very effective method for predicting regression problems. It has developed rapidly in recent years and has been widely used in different fields. Currently, artificial neural networks in machine learning are used in some fields of hearing aid gain, and machine learning has also begun to be used for the gain prediction of hearing aids.

[0004] Although there are currently few research methods for predicting hearing aid gain based on machine learning, in recent years, algorithms for predicting hearing aid gain based on DNN and those based on artificial neural networks have emerged. In clinical research, it has been found that they still have certain deficiencies: the gains during fitting are all based on one environment. In other environments, the gains are not appropriate, resulting in the phenomenon of "not being able to hear clearly at loud sounds and not being able to hear at soft sounds", and they cannot well simulate the adaptive changes of the hearing aid gain of the human ear based on different environments, thus leading to problems such as low prediction accuracy and insufficient generalization performance. Summary of the Invention

[0005] In view of the above problems, the object of the present invention is to provide a multi-environment hearing aid gain prediction method, system, processing device and storage medium with high prediction accuracy and high generalization performance.

[0006] To achieve the above object, the present invention adopts the following technical solutions: In the first aspect, a multi-environment hearing aid gain prediction method is provided, including: Obtain the audiogram of the patient; Input different hearing thresholds, different input sound pressure levels and environments in the audiogram of the patient into the trained multi-environment hearing aid gain prediction model to obtain the predicted results of the prescription gain at different frequencies corresponding to the patient, wherein the multi-environment hearing aid gain prediction model is constructed based on the XGBoost algorithm, grid search method and genetic algorithm.

[0007] Further, the construction process of the multi-environment hearing aid gain prediction model is as follows: Obtain the real data of several patients and process it, and construct a data set based on the hearing aid gain under different environments, wherein the real data of the patient includes the gender of each patient and the audiograms of the left and right ears; Adopt the GS-GA-XGBoost algorithm, and use the data set constructed with the hearing aid gain under different environments to construct and train a multi-environment hearing aid gain prediction model.

[0008] Further, the obtaining the real data of several patients and processing it, and constructing a data set based on the hearing aid gain under different environments includes: Obtain the real data of several patients; Input the audiogram in the real data of each patient into the fitting software to obtain the hearing thresholds at different frequency points and the prescription gains based on different environments under different input sound pressure levels, and construct a data set; Divide the constructed data set into a training set and a test set.

[0009] Further, the GS-GA-XGBoost algorithm is adopted, and a multi-environment hearing aid gain prediction model is constructed and trained using a dataset constructed with hearing aid gains in different environments, including: Adopt the XGBoost algorithm and construct a preliminary multi-environment hearing aid gain prediction model using a dataset constructed with hearing aid gains in different environments; Adopt the grid search method to determine the optimal values of the integer hyperparameters in the XGBoost algorithm and determine the optimal range of the floating-point hyperparameters in the XGBoost algorithm; Adopt the genetic algorithm to determine the optimal values of the floating-point hyperparameters based on the optimal range of the floating-point hyperparameters in the XGBoost algorithm; Optimize the multi-environment hearing aid gain prediction model according to the optimal values of the integer hyperparameters and the optimal values of the floating-point hyperparameters in the XGBoost algorithm to obtain the final multi-environment hearing aid gain prediction model.

[0010] Further, the adoption of the genetic algorithm to determine the optimal values of the floating-point hyperparameters based on the optimal range of the floating-point hyperparameters in the XGBoost algorithm includes: ① Adopt the genetic algorithm to initialize the population within the determined optimal range of the floating-point hyperparameters and set the genetic parameters of the population, where the genetic parameters include the crossover probability, mutation probability, population size, and number of iterations; ② Use population reconstruction. Based on the multi-environment hearing aid gain prediction model at this time, determine the fitness value of the individuals in the population as the RMSE of the multi-environment hearing aid gain prediction model; ③ Select the optimal individuals in the population for mutation, crossover, and genetic operations to generate offspring individuals, that is, the values of the new floating-point hyperparameters; ④ Optimize the preliminary multi-environment hearing aid gain prediction model based on the values of the new floating-point hyperparameters; ⑤ Determine whether the number of iterations has reached the number of iterations in the genetic parameters of the set population. If so, output the offspring individual when the RMSE of the multi-environment hearing aid gain prediction model is the smallest at this time as the optimal individual to obtain the optimal values of the floating-point hyperparameters; otherwise, go to step ②.

[0011] Further, the inputs of the multi-environment hearing aid gain prediction model are the different hearing thresholds, input sound pressure levels, and environments of the patient, and the output of the multi-environment hearing aid gain prediction model is the prescribed gain at different frequencies.

[0012] Further, the input sound pressure levels include 50 dB SPL and 80 dB SPL, and the environments include music environment, outdoor environment, restaurant environment, and traffic environment.

[0013] In a second aspect, a multi-environment hearing aid gain prediction system is provided, including: A data acquisition and determination module for acquiring the audiogram of a patient; A prescription gain prediction module for inputting different hearing thresholds, different input sound pressure levels, and the environment in the audiogram of the patient into a trained multi-environment hearing aid gain prediction model to obtain the prescription gain prediction results at different frequencies corresponding to the patient. Among them, the multi-environment hearing aid gain prediction model is constructed based on the XGBoost algorithm, the grid search method, and the genetic algorithm.

[0014] In a third aspect, a processing device is provided, including computer program instructions. When the computer program instructions are executed by the processing device, they are used to implement the steps corresponding to the above multi-environment hearing aid gain prediction method.

[0015] In a fourth aspect, a computer-readable storage medium is provided. A computer program instruction is stored on the computer-readable storage medium. When the computer program instruction is executed by a processor, it is used to implement the steps corresponding to the above multi-environment hearing aid gain prediction method.

[0016] Due to the above technical solutions adopted by the present invention, it has the following advantages: 1. The present invention can well solve the problem that the current hearing aid fitting is only based on one environment, resulting in the problems of "unable to hear clearly at high volume and unable to hear at low volume" when hearing-impaired patients wear hearing aids in other environments, realizing the adaptive change of hearing aid gain in different environments, with high prediction accuracy and strong generalization performance.

[0017] 2. The present invention uses the extreme gradient boosting (GS-GA-XGBoost) algorithm optimized by the grid search method and the genetic algorithm for hearing aid gain prediction in different environments, and quotes the gains of hearing aids in different environments at each frequency. The grid search method (GS) can quickly determine the optimal value and the range of the optimal value of integer hyperparameters. Using floating-point hyperparameters and the genetic algorithm (GA), the optimal value of floating-point hyperparameters can be quickly determined on the basis of the above work.

[0018] 3. The XGBoost algorithm optimized by the grid search method and the genetic algorithm in the present invention can quickly realize the construction of a hearing aid gain prediction model based on different environments. The constructed model has high accuracy in hearing aid gain prediction. When predicting the gain, each frequency band is independent, reducing the probability that the error of the entire frequency band becomes larger due to the error of a certain frequency band.

[0019] In summary, the present invention can be widely applied to the field of gain compensation for hearing aid fitting. Description of the Drawings

[0020] Upon reading the following detailed description of the preferred embodiments, various other advantages and benefits will become apparent to those of ordinary skill in the art. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not considered to be a limitation of the present invention. Throughout the drawings, the same reference numerals are used to denote the same components. In the drawings: Figure 1 is a schematic flow chart of a method provided by an embodiment of the present invention; Figure 2 is a schematic diagram for predicting the small and large sound gains of a hearing aid at each frequency point provided by an embodiment of the present invention, where Figure 2 (a) is a schematic diagram for predicting the small sound gain, Figure 2 (b) is a schematic diagram for predicting the large sound gain; Figure 3 is a schematic diagram for predicting the hearing aid gain at each frequency point in different environments provided by an embodiment of the present invention, where Figure 3 (a) is the optimal small sound model using the GS-GA-XGBoost algorithm, Figure 3 (b) is the optimal small sound model using the XGBoost algorithm, Figure 3 (c) is the optimal large sound model using the GS-GA-XGBoost algorithm, Figure 3 (d) is the optimal large sound model using the XGBoost algorithm. Detailed Embodiments

[0021] The exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present invention are shown in the drawings, it should be understood that the present invention can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided so that the present invention can be more thoroughly understood and the scope of the present invention can be fully communicated to those skilled in the art.

[0022] It should be understood that the terms used herein are for the purpose of describing specific exemplary embodiments only and are not intended to be limiting. Unless the context clearly indicates otherwise, the singular forms "a", "an", and "the" as used herein may also include the plural forms. The terms "comprising", "including", "containing", and "having" are inclusive and thus specify the presence of the stated features, steps, operations, elements, and / or components, but do not preclude the presence or addition of one or more other features, steps, operations, elements, components, and / or combinations thereof. The method steps, processes, and operations described herein are not to be construed as necessarily requiring them to be performed in the particular order described or illustrated, unless the order of performance is explicitly stated. It should also be understood that additional or alternative steps may be used.

[0023] Although terms such as first, second, third, etc. may be used in the text to describe multiple elements, components, regions, layers, and / or sections, these elements, components, regions, layers, and / or sections should not be limited by these terms. These terms may be used only to distinguish one element, component, region, layer, or section from another region, layer, or section. Unless the context clearly indicates otherwise, terms such as "first", "second", and other numerical terms do not imply order or sequence when used in the text. Therefore, the first element, component, region, layer, or section discussed below may be referred to as the second element, component, region, layer, or section without departing from the teachings of the exemplary embodiments.

[0024] Currently, the prior art discloses algorithms for predicting hearing aid gain based on DNN and prediction based on artificial neural networks. However, it has been found in clinical studies that they still have certain deficiencies: the gains during fitting are all based on one environment, and in other environments, the gains are not appropriate, resulting in the phenomenon of "not being able to hear clearly in loud sounds and not being able to hear in soft sounds", and it cannot well simulate the adaptive changes of the hearing aid gain based on different environments of the human ear. Furthermore, problems such as low prediction accuracy and insufficient generalization performance occur. The present invention needs to select a model with strong prediction accuracy, generalization performance, adaptability, and short prediction time. The XGBoost (Extreme Gradient Boosting) algorithm is an excellent method in machine learning. Compared with the gradient boosting decision tree algorithm, it has higher prediction accuracy, computing efficiency, and stronger generalization performance. The XGBoost algorithm has good performance and can perform parallel computing with multiple threads, and has high computing efficiency. However, the XGBoost algorithm without parameter optimization has a low fitting degree with the existing data set, resulting in poor generalization performance and adaptability. Therefore, the present invention uses the grid search method (GS) and genetic algorithm (GA) to optimize the hyperparameters of the XGBoost algorithm, and constructs an extreme gradient boosting algorithm (GS-GA-XGBoost) optimized based on the grid search method and genetic algorithm for predicting hearing aid gain values in different environments, where: 1. The grid search method (GS) is an exhaustive search method for determining given parameter values. It mainly combines every possible parameter value, generates a grid of these possible results, and then puts it into the training of the machine learning model. The fitting function will try all parameter combinations and finally return a classifier, and the classifier is automatically adjusted to the best parameter combination of the machine learning model.

[0025] 2. The genetic algorithm (GA) is a modern intelligent algorithm that simulates the survival of the fittest and natural genetic mechanisms in the biological world. The genetic algorithm starts from a group of randomly generated solutions or parameters to be optimized, and this group of solutions or parameters is a population; each individual in the population is a solution to the problem, and the individuals in the population are also chromosomes. By using the fitness function to evaluate the fitness of each individual in each generation, the higher the fitness of the individual, the higher the survival probability. The genetic algorithm selects the individuals with the highest fitness values for operations such as genetic crossover and mutation. The new population generated will be more in line with the objective function, forming a new population. After multiple iterations, it converges to the optimal solution until the optimal solution individual is generated or the set maximum number of iterations is reached and then it ends.

[0026] 3. The gradient boosting decision tree (XGBoost) algorithm. Its principle is to use the method of gradient descent to generate a new tree based on all the previous trees and make the objective function as small as possible. XGBoost is called extreme gradient boosting. It is a tree ensemble model that can be used to solve classification and regression problems. When it is used to solve regression problems, during the continuous addition of regression trees, the newly generated classification and regression tree (CART tree) will fit the residuals generated by the previous model. K represents all the tree numbers in the model, and the sum of the results corresponding to each tree is used as the final prediction value.

[0027] The main idea of the grid search method is that within the specified range, the parameters are adjusted according to the step size respectively. Therefore, this method is not ideal for the optimization of floating-point parameters. The genetic algorithm has a relatively fast convergence speed. When generating offspring, it can converge to the optimal solution through continuous genetic crossover and mutation operations, saving a considerable amount of time. Therefore, for the GS-GA-XGBoost algorithm proposed in the present invention, first, the grid search method is used to optimize the integer hyperparameters of the XGBoost algorithm to determine the optimization range of the floating-point hyperparameters, and then the genetic algorithm is used to obtain the optimal values of these floating-point hyperparameters. There are 5 hyperparameters in the XGBoost algorithm, namely: max_depth, min_child_weight, gamma, subsample, and colsample_bytree. The specific values are shown in Table 1 below: Table 1: Meanings of Hyperparameters in the XGBoost Algorithm Hyperparameter Type Meaning max_depth Integer The maximum depth of the tree min_child_weight Integer The minimum sum of instance weights of all samples in a leaf node gamma Float The minimum loss reduction required to further partition at the leaf nodes of the tree subsample Float The sampling rate of training samples colsample_bytree Float The column sampling rate of features when building each tree An embodiment of the present invention provides a multi - environment hearing aid gain prediction method, including: obtaining the audiogram of a patient; inputting different hearing thresholds, different input sound pressure levels, and the environment in the audiogram of the patient into a trained multi - environment hearing aid gain prediction model to obtain the prescription gain prediction results at different frequencies corresponding to the patient; wherein, the multi - environment hearing aid gain prediction model is constructed based on the XGBoost algorithm, grid search method, and genetic algorithm. The present invention can well solve the problem that current hearing aid fitting is only based on one environment, resulting in the situation that hearing - impaired patients wearing hearing aids have problems such as "unable to hear clearly at high volume and unable to hear at low volume" in other environments, realizing the adaptive change of hearing aid gain in different environments, with high prediction accuracy and strong generalization performance.

[0028] Embodiment 1 As Figure 1 shown, this embodiment provides a multi - environment hearing aid gain prediction method, including the following steps: 1) Obtain and process the real - world data of 150 patients, and construct a data set based on the hearing aid gain in different environments, specifically: 1.1) Obtain the real - world data of 150 patients, where the real - world data includes the gender of each patient and the audiograms of the left and right ears, etc.

[0029] 1.2) Input the audiograms in the real - world data of each patient into the fitting software to obtain the hearing thresholds at different frequency points, and the prescription gains based on different environments (such as: music, outdoors, restaurant, traffic) at different input sound pressure levels (for example: 50 dB SPL, 80 dB SPL), and construct a data set including 150 patients.

[0030] Specifically, the fitting software can be, for example, the fitting software of Rui** (R***), *Li (P**), or **Kang (O**).

[0031] Specifically, each patient in the data set includes data for both the left ear and the right ear, including the prescription gains at their respective low - volume (50 dBSPL) and high - volume (80 dB SPL) input sound pressure levels. Therefore, a total of 150×2×2 = 600 samples are obtained. Among them, each sample includes the hearing thresholds, input sound pressure levels, and four environments (music environment, outdoor environment, restaurant environment, traffic environment) at 6 frequency points of 250, 500, 1000, 2000, 4000, 8000 Hz for the patient, as well as the prescription gains at 9 frequencies of 250, 500, 750, 1000, 1500, 2000, 3000, 4000, 6000 Hz, for a total of 150×2×2×9×4 = 21600 samples.

[0032] 1.3) Divide the constructed data set into a training set and a test set.

[0033] Specifically, in the present invention, different hearing thresholds, input sound pressure levels, and environments of patients are selected as the inputs of the prediction model, and the prescription gains at different frequencies are used as the outputs of the prediction model. In this embodiment, the order of the existing dataset is shuffled and reorganized to form a training set and a test set (90% of the dataset is used as the training set for learning the GS-GA-XGBoost algorithm, and the subsequent 10% is used as the test set for prediction). Among them, the first column is the id name of different ears; the second, third, fourth, and fifth columns correspond to different environments, namely music environment, outdoor environment, restaurant environment, and traffic environment; the sixth to eleventh columns correspond to the hearing thresholds of patients, which are the hearing thresholds of patients at 250, 500, 1000, 2000, 4000, and 8000 Hz respectively; the twelfth column corresponds to the frequency points, and the content therein includes the frequencies of 9 channels, which are 250, 500, 750, 1000, 1500, 2000, 3000, 4000, and 6000 Hz respectively; the thirteenth and fourteenth columns are the prescription gains at the corresponding frequency points when the input sound pressure levels are soft and loud respectively.

[0034] 2) Adopt the GS-GA-XGBoost algorithm, and use the dataset constructed by the hearing aid gains in different environments to construct and train a multi-environment hearing aid gain prediction model, specifically: 2.1) Adopt the XGBoost algorithm, and use the dataset constructed by the hearing aid gains in different environments to construct a preliminary multi-environment hearing aid gain prediction model.

[0035] Specifically, the inputs of the multi-environment hearing aid gain prediction model are different hearing thresholds, input sound pressure levels, and environments of patients, and the outputs of the multi-environment hearing aid gain prediction model are the prescription gains at different frequencies.

[0036] Specifically, the 5 hyperparameters of the XGBoost algorithm are all set to default values.

[0037] 2.2) Adopt the grid search method to determine the optimal values of the integer hyperparameters max_depth and min_child_weight in the XGBoost algorithm, and determine the optimal ranges of the floating-point hyperparameters gamma, subsample, and colsample_bytree in the XGBoost algorithm.

[0038] 2.3) Adopt the genetic algorithm to determine the optimal values of the floating-point hyperparameters based on the optimal ranges of the floating-point hyperparameters in the XGBoost algorithm: 2.3.1) Adopt the genetic algorithm to initialize the population within the determined optimal ranges of the floating-point hyperparameters, and set the genetic parameters of the population, where the genetic parameters include the crossover probability, mutation probability, population size, and number of iterations.

[0039] 2.3.2) Using population reconstruction, based on the multi-environment hearing aid gain prediction model at this time, determine the fitness value of the individuals in the population as the RMSE (Root Mean Squared Error) of the multi-environment hearing aid gain prediction model.

[0040] 2.3.3) Select the optimal individual in the population for mutation, crossover, and genetic operations to generate offspring individuals, that is, the values of the new floating-point hyperparameters.

[0041] 2.3.4) Optimize the preliminary multi-environment hearing aid gain prediction model based on the values of the new floating-point hyperparameters.

[0042] 2.3.5) Determine whether the number of iterations has reached the number of iterations in the genetic parameters of the set population. If so, output the offspring individual with the smallest RMSE of the multi-environment hearing aid gain prediction model at this time as the optimal individual to obtain the optimal value of the floating-point hyperparameters; otherwise, go to step 2.3.2).

[0043] 2.4) Optimize the multi-environment hearing aid gain prediction model according to the optimal values of the integer hyperparameters and the optimal values of the floating-point hyperparameters in the XGBoost algorithm to obtain the final multi-environment hearing aid gain prediction model.

[0044] 2.5) Based on the training set and the test set, train and test the final multi-environment hearing aid gain prediction model to obtain a trained multi-environment hearing aid gain prediction model.

[0045] 3) Obtain the audiogram of the patient.

[0046] 4) Input the different hearing thresholds, different input sound pressure levels, and the environment in the audiogram of the patient into the trained multi-environment hearing aid gain prediction model to obtain the prescription gain prediction results at different frequencies corresponding to the patient. By using the method of the present invention for hearing aid gain prediction, the problem that the current hearing aid gain only exists in one environment can be solved and the prediction accuracy of the model is relatively high.

[0047] The following takes the XGBoost algorithm and the GS-XGBoost algorithm as comparative examples to illustrate in detail the multi-environment hearing aid gain prediction method of the present invention: As Figure 2 shown, the difference between the predicted value and the actual value of the GS-GA-XGBoost algorithm of the present invention at each frequency is basically about 0.5 dB, while the step size of the actual hearing aid gain adjustment is greater than 1 dB. Generally, by default, as long as the error from the actual value does not exceed 5 dB, the default effect is considered good.

[0048] As Figure 3As shown, based on Figure 2 the prediction effects of the GS-GA-XGBoost algorithm and the XGBoost algorithm on samples in different environments are presented. It can be seen that the best-performing model is the GS-GA-XGBoost algorithm of the present invention, while the worst-performing is the XGBoost algorithm. The abscissa in the figure is the frequency at the corresponding channel during prediction, and the ordinate is the mean of the differences between the actual values and the predicted values of all tested patients. The error bars on the columns in the bar chart represent the variances of the differences between the actual values and the predicted values of all tested patients at this frequency in the corresponding environment. From Figure 3 Figures (a) and (c), it can be seen that under the optimal models for soft voice and loud voice, the average errors in the four environments are relatively small. Among them, the errors of the loud voice optimal model at different frequencies in different environments are significantly smaller than those of the soft voice optimal model. The mean of the differences between its actual value and the predicted value is mostly around 0.5 dB, with a maximum not exceeding 1 dB, and the variances of the errors at each frequency are also relatively small, and the errors after prediction are very stable. From Figure 3 Figures (b) and (d), it can be seen that under the worst models for soft voice and loud voice, the average errors in the four environments increase significantly compared to the optimal model (GS-GA-XGBoost). In the worst model for soft voice (XGBoost), as Figure 3 shown in Figure (b), the average error of the difference between the actual value and the predicted value of the gain in some environments has reached about 3 dB. The average error of the worst model for loud voice (XGBoost) also reaches nearly 2 dB, and the variances of the differences between the actual value and the predicted value of the gain in many environments and frequencies are also very large, and the predicted errors are very unstable. Comparing Figure 3 Figures (b) and (d), the overall errors of the worst model for loud voice are smaller than those of the worst model for soft voice. The main reason is that the soft voice gain value is much larger than the loud voice gain value during prediction, and the range of the soft voice gain value is wider. Therefore, when predicting the soft voice gain, the error will be larger. Generally speaking, the GS-GA-XGBoost algorithm of the present invention has good prediction effects at different frequencies in different environments, and the errors are relatively stable. Whether it is the soft voice model or the loud voice model, the mean of the differences between the actual value and the predicted value does not exceed 1 dB. At present, the minimum step size of the gain adjustment in the hearing aid fitting software is also 1 dB. In practice, as long as the actual error does not exceed 5 dB, the prediction effect can basically be considered good. Therefore, the prediction effect of the method of the present invention basically meets the actual requirements.

[0049] Example 2 This example provides a multi-environment hearing aid gain prediction system, including: A data acquisition and determination module for acquiring the audiogram of a patient.

[0050] A prescription gain prediction module, which is configured to input different hearing thresholds, different input sound pressure levels, and the environment in the audiogram of the patient into a trained multi-environment hearing aid gain prediction model to obtain the prescription gain prediction results at different frequencies corresponding to the patient.

[0051] Among them, the multi-environment hearing aid gain prediction model is constructed based on the XGBoost algorithm, the grid search method, and the genetic algorithm.

[0052] The system provided in this embodiment is used to execute the above method embodiments. For the specific process and detailed content, please refer to the above embodiments and will not be elaborated here.

[0053] Embodiment 3 This embodiment provides a processing device corresponding to the multi-environment hearing aid gain prediction method provided in Embodiment 1. The processing device can be a processing device applicable to a client, such as a mobile phone, a laptop computer, a tablet computer, a desktop computer, etc., to execute the method of Embodiment 1.

[0054] The processing device includes a processor, a memory, a communication interface, and a bus. The processor, the memory, and the communication interface are connected through the bus to complete communication with each other. A computer program that can run on the processing device is stored in the memory. When the processing device runs the computer program, it executes the multi-environment hearing aid gain prediction method provided in Embodiment 1 of this embodiment.

[0055] In some implementations, the memory can be a high-speed random access memory (RAM: Random Access Memory), and may also include a non-volatile memory, such as at least one disk memory.

[0056] In other implementations, the processor can be a general-purpose processor of various types such as a central processing unit (CPU) and a digital signal processor (DSP), which are not limited here.

[0057] In addition, when the logical instructions in the above-mentioned memory are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on such understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The foregoing storage medium includes: USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs, etc., all kinds of media that can store program codes.

[0058] Those skilled in the art can understand that the structure of the above-mentioned computing device is only a part of the structure related to the solution of the present invention, and does not constitute a limitation on the computing device to which the solution of the present invention is applied. The specific computing device may include more or fewer components, or combine some components, or have different component arrangements.

[0059] Embodiment 4 This embodiment provides a computer program product corresponding to the multi-environment hearing aid gain prediction method provided in Embodiment 1. The computer program product may include a computer-readable storage medium, on which computer-readable program instructions for executing the multi-environment hearing aid gain prediction method described in Embodiment 1 are loaded.

[0060] A computer-readable storage medium may be a tangible device that holds and stores instructions used by an instruction execution device. A computer-readable storage medium may be, for example, but not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any combination of the above.

[0061] The computer-readable storage medium provided in the above embodiment has the same implementation principle and technical effect as the above method embodiment, and will not be elaborated here.

[0062] The present invention is described with reference to the flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each flow and / or block in the flowchart and / or block diagram, and the combination of flows and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices are used to implement the processFigure 1 means for the functions specified in one process or more processes and / or one block or more blocks Figure 1 or more blocks.

[0063] These computer program instructions can also be stored in a computer-readable memory capable of guiding a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured article including an instruction device, and the instruction device implements the process Figure 1 means for the functions specified in one process or more processes and / or one block or more blocks Figure 1 or more blocks.

[0064] These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process, and thus the instructions executed on the computer or other programmable device provide for implementing the process Figure 1 means for the functions specified in one process or more processes and / or one block or more blocks Figure 1 or more blocks.

[0065] The above embodiments are only used to illustrate the present invention. The structures, connection methods, manufacturing processes, etc. of the components can all be changed. Any equivalent transformation and improvement based on the technical solution of the present invention should not be excluded from the protection scope of the present invention.

Claims

1. A method for predicting hearing aid gain in multiple environments, characterized in that: include: Obtain the patient's audiogram; The different hearing thresholds in the patient's audiogram and different input sound pressure levels and environments are input into a trained multi-environment hearing aid gain prediction model to obtain the prescription gain prediction results at different frequencies corresponding to the patient, wherein the multi-environment hearing aid gain prediction model is constructed based on the XGBoost algorithm, the grid search method and the genetic algorithm.

2. The method for predicting hearing aid gain in multiple environments according to claim 1, characterized in that: The construction process of the multi-environment hearing aid gain prediction model is as follows: Obtain and process the real data of several patients, and construct a data set based on the hearing aid gain in different environments. The real data of the patients includes the gender of each patient and the audiograms of the left and right ears. The GS-GA-XGBoost algorithm was used to construct and train a multi-environment hearing aid gain prediction model using a dataset constructed with hearing aid gains in different environments.

3. The method for predicting hearing aid gain in multiple environments as claimed in claim 2, characterized in that: The real data of several patients are obtained and processed, and a data set is constructed based on the hearing aid gain in different environments, including: Obtain real data from several patients; Input the audiogram of each patient's real data into the fitting software to obtain the hearing thresholds at different frequencies and the prescription gain based on different environments at different input sound pressure levels, and construct a data set; The constructed dataset is divided into a training set and a test set.

4. The method for predicting hearing aid gain in multiple environments as claimed in claim 3, characterized in that: The GS-GA-XGBoost algorithm is used to construct and train a multi-environment hearing aid gain prediction model using a data set constructed using hearing aid gains in different environments, including: The XGBoost algorithm was used to construct a preliminary multi-environment hearing aid gain prediction model using a dataset constructed from hearing aid gains in different environments. Use grid search method to determine the optimal value of integer hyperparameters in XGBoost algorithm, and determine the optimal range of floating point hyperparameters in XGBoost algorithm; Using genetic algorithm, based on the optimal range of floating point hyperparameters in XGBoost algorithm, the optimal value of floating point hyperparameters is determined; According to the optimal values ​​of the integer hyperparameters and the optimal values ​​of the floating-point hyperparameters in the XGBoost algorithm, the multi-environment hearing aid gain prediction model is optimized to obtain the final multi-environment hearing aid gain prediction model.

5. The method for predicting hearing aid gain in multiple environments as claimed in claim 4, characterized in that: The method of using a genetic algorithm to determine the optimal value of a floating point hyperparameter based on the optimal range of the floating point hyperparameter in the XGBoost algorithm includes: ① Using genetic algorithm, the population is initialized within the optimal range of the determined floating point hyperparameters, and the genetic parameters of the population are set, where the genetic parameters include crossover probability, mutation probability, population size, and number of iterations; ② Using population reconstruction, based on the multi-environment hearing aid gain prediction model at this time, the fitness value of the individual in the population is determined as the RMSE of the multi-environment hearing aid gain prediction model; ③ Select the best individuals in the population for mutation, crossover and genetic operations to produce offspring individuals, that is, the values ​​of new floating-point hyperparameters; ④ Optimize the preliminary multi-environment hearing aid gain prediction model based on the values ​​of the new floating-point hyperparameters; ⑤ Determine whether the number of iterations reaches the number of iterations in the set genetic parameters of the population. If so, output the offspring individual with the smallest RMSE of the multi-environment hearing aid gain prediction model as the optimal individual to obtain the optimal value of the floating-point hyperparameter; otherwise, go to step ②.

6. The method for predicting hearing aid gain in multiple environments as claimed in claim 2, characterized in that: The input of the multi-environment hearing aid gain prediction model is different hearing thresholds of patients, input sound pressure levels and environments, and the output of the multi-environment hearing aid gain prediction model is prescription gains at different frequencies.

7. The method for predicting hearing aid gain in multiple environments according to claim 6, characterized in that: The input sound pressure levels include 50 dB SPL and 80 dB SPL, and the environments include a music environment, an outdoor environment, a restaurant environment, and a traffic environment.

8. A multi-environment hearing aid gain prediction system, characterized in that: include: A data acquisition determination module is used to obtain the patient's audiogram; The prescription gain prediction module is used to input different hearing thresholds in the patient's audiogram and different input sound pressure levels and environments into a trained multi-environment hearing aid gain prediction model to obtain the prescription gain prediction results at different frequencies corresponding to the patient, wherein the multi-environment hearing aid gain prediction model is constructed based on the XGBoost algorithm, the grid search method and the genetic algorithm.

9. A processing device, characterized in that: The method comprises computer program instructions, wherein when the computer program instructions are executed by a processing device, they are used to implement the steps corresponding to the multi-environment hearing aid gain prediction method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer program instructions, wherein the computer program instructions are used to implement the steps corresponding to the multi-environment hearing aid gain prediction method according to any one of claims 1 to 7 when executed by a processor.

Images