CN119421097B - Conversion method of multiple hearing aid fitting parameters based on clinical auditory sense supervision - Google Patents
Conversion method of multiple hearing aid fitting parameters based on clinical auditory sense supervision Download PDFInfo
- Publication number
- CN119421097B CN119421097B CN202510019428.9A CN202510019428A CN119421097B CN 119421097 B CN119421097 B CN 119421097B CN 202510019428 A CN202510019428 A CN 202510019428A CN 119421097 B CN119421097 B CN 119421097B
- Authority
- CN
- China
- Prior art keywords
- hearing aid
- hearing
- parameters
- fitting
- test
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04R—LOUDSPEAKERS, MICROPHONES, GRAMOPHONE PICK-UPS OR LIKE ACOUSTIC ELECTROMECHANICAL TRANSDUCERS; DEAF-AID SETS; PUBLIC ADDRESS SYSTEMS
- H04R25/00—Deaf-aid sets, i.e. electro-acoustic or electro-mechanical hearing aids; Electric tinnitus maskers providing an auditory perception
- H04R25/30—Monitoring or testing of hearing aids, e.g. functioning, settings, battery power
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04R—LOUDSPEAKERS, MICROPHONES, GRAMOPHONE PICK-UPS OR LIKE ACOUSTIC ELECTROMECHANICAL TRANSDUCERS; DEAF-AID SETS; PUBLIC ADDRESS SYSTEMS
- H04R2225/00—Details of deaf aids covered by H04R25/00, not provided for in any of its subgroups
- H04R2225/41—Detection or adaptation of hearing aid parameters or programs to listening situation, e.g. pub, forest
Landscapes
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Otolaryngology (AREA)
- Neurosurgery (AREA)
- Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Acoustics & Sound (AREA)
- Signal Processing (AREA)
- Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)
Abstract
本发明公开了一种基于临床听感监督下的多种助听器验配参数的转换方法，具体涉及助听器验配参数转换技术领域，该方法包括：步骤一、构建用户特征、助听器特征，步骤二、生成一级各助听器的验配参数，步骤三、生成统一的隐变量，步骤四、生成转换后助听器验配参数，步骤五、验证助听器验配参数转换模型。本发明采用用户特征和各助听器的特征作为验配参数转换模型的输入端数据，避免了在助听器验配参数转换过程中依赖于听损患者的反馈，根据用户特征和各助听器特征可以一次性输出不同助听器的同听感验配参数，同时对验配参数转换模型进行微调训练，使验配参数转换模型具有自监督学习能力，提高验配参数转换模型的稳定性。
The present invention discloses a method for converting a plurality of hearing aid fitting parameters under clinical hearing supervision, and specifically relates to the technical field of hearing aid fitting parameter conversion, the method comprising: step one, constructing user features and hearing aid features, step two, generating fitting parameters of each hearing aid at the first level, step three, generating a unified latent variable, step four, generating converted hearing aid fitting parameters, and step five, verifying the hearing aid fitting parameter conversion model. The present invention adopts user features and features of each hearing aid as input data of the fitting parameter conversion model, avoiding reliance on feedback from hearing-impaired patients during the hearing aid fitting parameter conversion process, and can output the same hearing sense fitting parameters of different hearing aids at one time according to user features and features of each hearing aid, and at the same time, fine-tunes the fitting parameter conversion model, so that the fitting parameter conversion model has self-supervised learning capability, and improves the stability of the fitting parameter conversion model.
Description
Technical Field
The invention relates to the technical field of hearing aid fitting parameter conversion, in particular to a conversion method of multiple hearing aid fitting parameters based on clinical hearing supervision.
Background
Hearing aids of different appearance types and product families typically employ different hearing aid chips, which also correspond to different parameter systems. There is no direct conversion tool between the chip parameters, and even if the same hearing aid chip is used, the same hearing aid parameters are configured, and due to the difference of physical characteristics, the hearing feeling of the user is also different, and the user can use the hearing aid for a longer adaptation period. For a daily user change scenario, hearing impaired patients always want to be able to restore the hearing sensation of the hearing aid before the change to avoid lengthy adaptation processes. Thus, a study of a variety of hearing aid fitting parameter switching methods is necessary.
The invention discloses an all-digital hearing aid and a non-traditional hearing aid fitting method, wherein the method comprises the steps of presetting at least two hearing compensation prescription modes which are subjected to corresponding hearing aid parameter programming or non-linear signal processing aiming at typical non-linear hearing loss diagram types of target hearing impaired patients, enabling hearing impaired patients to wear the hearing aid directly without professional fitting services of traditional digital hearing aids, including but not limited to hearing test and hearing aid programming debugging, selecting a hearing compensation prescription mode which is suitable for hearing loss types and hearing loss degrees of hearing impaired patients, adjusting volume in the hearing compensation prescription mode, and selecting the hearing compensation mode and volume according to self-perception hearing compensation effect, wherein the all-digital hearing aid disclosed by the invention can meet the self-fitting requirements of hearing impaired patients, and has wide application range.
The prior art discloses a hearing aid fitting method, a device, computer equipment and a storage medium disclosed in the patent application of CN110719558B, wherein the method comprises the steps of obtaining voice response of a user to excitation sound, analyzing the voice response through semantic analysis to obtain response semantics, determining hearing loss of the user according to the excitation sound and the response semantics, determining hearing compensation corresponding to the user according to the hearing loss of the user, and configuring hearing aid parameters according to the hearing compensation. The method comprises the steps of releasing excitation sound, obtaining voice response of a user to the excitation sound, analyzing response semantics of the voice response, determining hearing loss of the user according to the excitation sound and the response semantics corresponding to the excitation sound, determining hearing compensation corresponding to the user according to the hearing loss, and configuring hearing aids corresponding to the user according to the hearing compensation pair.
According to the scheme, the fact that in the prior art, the conversion of the hearing aid fitting parameters is mostly dependent on professional fitters to adjust according to the parameters of the hearing aid before conversion, the hearing aid parameter indexes after conversion and own experience, and the feedback of hearing loss patients is adopted, so that on one hand, the conversion time of the hearing aid fitting parameters is increased, on the other hand, the accuracy of the hearing aid fitting parameter conversion is reduced due to the fact that the hearing aid fitting parameters are dependent on own experience of the fitters and the feedback of the hearing loss patients, the requirements of the hearing loss patients cannot be met after the hearing aid fitting parameters are actually converted, and the hearing compensation and programming technology is adopted to adjust the hearing aid fitting parameters, but the hearing aid fitting parameters are still dependent on the feedback of the hearing loss patients in the process, so that the hearing aid fitting parameter conversion efficiency is low, the accuracy is low, and the practical application requirements are difficult to meet in the practical use process.
Disclosure of Invention
The invention aims to provide a conversion method of various hearing aid fitting parameters based on clinical hearing supervision, which solves the problems in the background technology.
In order to solve the technical problems, the invention adopts the following technical scheme that the method for converting the test and matching parameters of a plurality of hearing aids based on clinical hearing supervision comprises the following steps:
Step one, constructing user characteristics and hearing aid characteristics, namely acquiring basic information and hearing information of a user, constructing user characteristics, acquiring information of all hearing aids and constructing the characteristics of all hearing aids.
Step two, generating the fitting parameters of each first-level hearing aid based on the user characteristics byNetwork, generating hidden variable, inputting hidden variable and each hearing aid characteristicIn the network, fitting parameters of the first-class hearing aids are generated.
Step three, generating unified hidden variables, namely matching the characteristics of the hearing aid to be converted and the characteristics of the hearing aid after conversion from the characteristics of each hearing aid, and selecting the first-level hearing aid fitting parameters corresponding to the characteristics of the hearing aid to be converted and the characteristics of the hearing aid before conversion as conversionAnd (3) inputting the network to generate unified hidden variables.
Generating the test and matching parameters of the converted hearing aid, namely taking the characteristics of the converted hearing aid and the unified hidden variables as based on the unified hidden variablesThe inputs to the network generate fitting parameters for the transformed hearing aid.
Step five, verifying a hearing aid fitting parameter conversion model, namely selecting a test hearing aid from a databaseIs to test the hearing aidAfter the characteristics and the fitting parameters of the hearing aid are processed by a hearing aid fitting parameter conversion model, a test hearing aid is obtainedIs used for judging and testing the hearing aidAnd testing a hearing aidWhether the fitting parameters belong to the hearing-sensing fitting parameters or not, and performing fine tuning training.
Compared with the prior art, the invention has the beneficial effects that firstly, the user characteristics and the characteristics of each hearing aid are adopted as the input end data of the test-fit parameter conversion model, the feedback of a hearing loss patient in the process of the test-fit parameter conversion of the hearing aid is avoided, and the efficiency of the test-fit parameter conversion of the hearing aid is improved.
2. The invention adopts the deep neural network algorithm, can output the hearing-identical sensing and matching parameters of different hearing aids at one time according to the user characteristics and the hearing aid characteristics, is convenient for matching the matching parameters of the hearing aids after the conversion in the actual hearing aid matching parameter conversion process, and improves the reusability of the matching parameter conversion model.
3. According to the invention, according to the fitting parameters of a large number of hearing aids, the fitting parameters of the hearing aids of the patients with real hearing loss and the fitting parameters converted by the fitter, the hearing-aid hearing-sensing fitting parameters are obtained, and are used as the verification data of the fitting parameter conversion model, so that the accuracy of the fitting parameter conversion model is improved, and the requirements of users in the actual use process are met.
4. The invention analyzes the simultaneous hearing and sensing fitting parameters of each hearing aid and the fitting data converted by the actual hearing aid, performs fine tuning training on the fitting parameter conversion model, refines the accuracy of user hearing and sensing copying, ensures that the fitting parameter conversion model has self-supervision learning capability, and improves the stability of the fitting parameter conversion model.
Drawings
In order to more clearly illustrate the embodiments of the invention or the technical solutions in the prior art, the drawings that are required in the embodiments or the description of the prior art will be briefly described, it being obvious that the drawings in the following description are only some embodiments of the invention, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a schematic flow chart of the method of the present invention.
Fig. 2 is a schematic diagram of a model structure of the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
For ease of understanding, the following terms are explained with respect to the present invention:
Refers to techniques commonly used in deep learning and neural networks that, in probability-based encoder-decoder architectures, The effect of (a) is to map the input to a distribution of potential space.
In a probability-based encoder-decoder architecture,The effect of which is to generate output data, typically data similar to the input data structure, from samples of the underlying space.
Fitting parameters, which are various audio processing parameters set according to individual hearing loss and requirements during fitting and adjustment of the hearing aid, help optimize the performance of the hearing aid, ensuring that the wearer has an optimal hearing experience.
Hidden variable means pass throughThe input data is mapped to a low-dimensional, highly abstract space that generally represents the core features or structure of the input data, rather than the original data itself. The dimensions of the potential space are typically smaller than those of the original data in order to capture the essential features of the data through the learned data representation.
Space dimension refers toThe vector or representation of the output is a number of dimensions in the potential space. This dimension determines the complexity and compactness of the representation of the data after it has been mapped from the input space to the potential space.
Hearing sensation refers to the habit degree of a person on sound, and the hearing sensation refers to the hearing sensation of a hearing impaired patient on a hearing aid to be converted as similar as possible to the hearing sensation of a hearing aid after conversion.
The back propagation algorithm is one of the most commonly used optimization algorithms when training the neural network, is an optimization method based on gradient descent, and is used for calculating the gradient of parameters (weight and bias) of each layer in the neural network, so that the performance of the test matching parameter conversion model is optimized through gradient descent, and the prediction error of the test matching parameter conversion model is reduced.
The learning rate of the network refers to the step length of the network weight adjustment when the parameters are updated each time in the deep learning and the neural network, and the learning rate determines how to adjust the parameters to minimize the loss function in the training process of the model.
Gradient operation, which is a partial derivative vector of a multi-variable function that represents the maximum rise direction of the function at a point, in machine learning, gradient generally refers to the calculation of the derivative of a loss function with respect to model parameters (e.g., weights and biases) by a back-propagation algorithm.
Embodiment one:
referring to fig. 1, a method for converting a plurality of hearing aid fitting parameters under clinical hearing supervision, the method comprising:
Step one, constructing user characteristics and hearing aid characteristics, namely acquiring basic information and hearing information of a user, constructing user characteristics, acquiring information of all hearing aids and constructing the characteristics of all hearing aids.
Step two, generating the fitting parameters of each first-level hearing aid based on the user characteristics byNetwork, generating hidden variable, inputting hidden variable and each hearing aid characteristicIn the network, fitting parameters of the first-class hearing aids are generated.
Step three, generating unified hidden variables, namely matching the characteristics of the hearing aid to be converted and the characteristics of the hearing aid after conversion from the characteristics of each hearing aid, and selecting the first-level hearing aid fitting parameters corresponding to the characteristics of the hearing aid to be converted and the characteristics of the hearing aid before conversion as conversionAnd (3) inputting the network to generate unified hidden variables.
Generating the test and matching parameters of the converted hearing aid, namely taking the characteristics of the converted hearing aid and the unified hidden variables as based on the unified hidden variablesThe inputs to the network generate fitting parameters for the transformed hearing aid.
Step five, verifying a hearing aid fitting parameter conversion model, namely selecting a test hearing aid from a databaseIs to test the hearing aidAfter the characteristics and the fitting parameters of the hearing aid are processed by a hearing aid fitting parameter conversion model, a test hearing aid is obtainedIs used for judging and testing the hearing aidAnd testing a hearing aidWhether the fitting parameters belong to the hearing-sensing fitting parameters or not, and performing fine tuning training.
In a specific embodiment, the database is used for storing the type and chip type of each hearing aid, the hearing aid hearing-aid sensing and fitting parameters, and the database is used for storingThe learning rate of the network,Is set to be a constant value.
In a specific embodiment of the invention, the method for constructing the user characteristics comprises the steps of extracting age and sex characteristic data of a user based on basic information input by the user in advance to obtain the basic information of the user.
Based on audiogram input by user, extracting hearing threshold value, type of hearing loss and grade of hearing loss of user to obtain hearing information of user.
In a specific embodiment, the user inputs basic information first, then inputs audiogram of the user, extracts hearing information of the user, and constructs user characteristics.
It should be noted that in a specific embodiment the influence of the sex of the user on the switching of the fitting parameters of the hearing aid comprises that a male typically has a more pronounced hearing loss in the high frequency range, e.g. 2kHz-8kHz, and thus a higher high frequency gain needs to be considered for the switching of the fitting parameters of the male hearing aid, and a female typically has a significant hearing loss in the medium and high frequency range, e.g. 1kHz-4kHz, and thus a more balanced gain needs to be considered for the switching of the fitting parameters of the female hearing aid, avoiding high frequency overamplification.
Older users experience high frequency hearing loss and cannot clearly hear high frequency sounds, so that the gain of the high frequency part needs to be considered when the hearing aid fitting parameters of older users are switched, the sensitivity of older users to volume changes, the higher volume or the excessively strong gain may cause discomfort, and the output limitation and volume control need to be paid attention to when the hearing aid fitting parameters of older users are switched so as to ensure the comfort of the sounds.
In a specific embodiment, the audiogram of the user is a chart for recording and evaluating the hearing ability of a person, and after the hearing test, the horizontal axis of the audiogram represents frequency, the units are hertz, the audiogram is arranged from low frequency to high frequency, the vertical axis of the audiogram represents loudness, the units are decibels, the range is usually 0dB to 120dB, the hearing threshold line marks the lowest volume which can be heard by a tester in the audiogram at each frequency, and the lower the hearing threshold value is, the better the hearing ability is, the higher the hearing threshold value is, and the more serious the hearing loss is.
Types of hearing loss include conductive hearing loss, which is typically manifested as a hearing threshold that is high at low frequencies and remains normal or slightly lost at high frequencies, sensorineural hearing loss, which is typically manifested as a hearing threshold that is high at both low and high frequencies and exhibits a smoother increase in threshold, and mixed hearing loss, which is typically manifested as a conductive loss at low frequencies and sensorineural loss at high frequencies.
The level of hearing loss is determined based on the interval in which the hearing threshold is located, e.g. the hearing threshold is locatedDetermining that the hearing loss is of a mild hearing loss level and the hearing threshold is atDetermining that the hearing loss is of a medium hearing loss level and the hearing threshold is inAnd judging the grade of the hearing loss as severe hearing loss, and judging the grade of the hearing loss as severe hearing loss if the hearing threshold is larger than 80.
Combining basic information and hearing information of a user to construct user characteristics, which are recorded as。
In a specific embodiment of the invention, the method for constructing the characteristics of each hearing aid comprises the steps of extracting the type and the chip type of each hearing aid from a database, combining the type and the chip type of each hearing aid to obtain information of each hearing aid, and constructing the characteristics of each hearing aid based on the information of each hearing aid, wherein the characteristics are recorded asWherein,The number of each hearing aid is indicated,Indicating the total number of hearing aids.
The types of hearing aids include behind-the-ear hearing aids, in-the-ear hearing aids, canal hearing aids, and bone conduction hearing aids.
The chip types of each hearing aid comprise a high-end signal processing chip, a medium-end signal processing chip, a low-end signal processing chip and an artificial intelligent chip.
It should be noted that in a specific embodiment, the high-end signal processing chip has the following characteristics of strong noise management and background noise filtering functions, and can provide clear voice hearing in a complex noise environment, for example Signia Xperience, has the functions of environment sensing, voice focusing and the like, and Phonak Belong supports voice definition and background noise management.
The middle-low end signal processing chip has the characteristics of having a certain background noise filtering function, being not as good as a high-end chip, being capable of automatically adjusting the volume in a mute and noise environment, but generally not providing highly accurate adjustment like the high-end chip, such as Oticon Siya, providing basic sound amplification and noise filtering functions, reSound ENZO D and supporting simple noise management and feedback suppression.
The artificial intelligent chip has the characteristics that the hearing environment of a wearer can be perceived in real time, the audio setting is automatically adjusted, the voice definition and the background noise suppression are optimized, the intelligent voice recognition and the background noise filtering are supported by the Phonak Marvel AI, the voice definition is improved by using a deep learning technology, and the environmental sound perception can be adjusted in real time, wherein Oticon More is realized.
The invention adopts the user characteristics and the characteristics of each hearing aid as the input end data of the test-fit parameter conversion model, avoids the feedback of a hearing loss patient in the process of the test-fit parameter conversion of the hearing aid, and improves the efficiency of the test-fit parameter conversion of the hearing aid.
In a specific embodiment of the invention, the method for generating the hidden variable comprises the following steps ofNetwork, willThe potential spatial dimension of the network is set to 128.
Based on the user characteristics, by model: characterizing the user As a means ofInput data of the network, throughEncoding process of network, outputting hidden variable, and recording asWhereinRepresentation ofProcessing of the network.
In a specific embodiment of the invention, the method for generating the fitting parameters of each first-level hearing aid comprises the following steps ofFeatures of the respective hearing aidsBy the model: Sequentially adding hidden variables And the firstFeatures of individual hearing aidsAs a means ofInput end data of the network, generating a first levelFitting parameters of individual hearing aidsWhereinRepresentation ofThe processing procedure of the network is that,Representing the input data connection symbols.
Similarly, the hidden variable and the characteristics of each hearing aid are sequentially taken asAnd obtaining the input end data of the network to obtain the fitting parameters of each first-level hearing aid.
In a specific embodiment, the fitting parameters of the hearing aid include gain, maximum output sound pressure level, frequency response, feedback suppression, noise suppression, maximum gain and sensitivity.
In a specific embodiment of the invention, the conversionThe network specifically analyzing method comprises the following steps of based onNetwork model: By inputting hidden variables and first level Features of individual hearing aids, viaAfter the decoding processing of the network, obtain the firstFitting parameters of individual hearing aids, willReverse reconfiguration of network through conversionNetwork model: Will be at the first Fitting parameters and characteristics of individual hearing aids as steeringInput data of the network, carrying out recovery of hidden variables, whereinRepresenting the reduced hidden variable, whereinIndicating steeringProcessing of the network.
In a specific embodiment of the invention, the unified hidden variables are generated by matching the characteristics of the hearing aids to be converted from the characteristics of each hearing aid, and marking the characteristics of the hearing aids to be converted as the characteristics of the hearing aids to be convertedWhereinMatching the fitting parameters of the first-level hearing aid corresponding to the characteristics of the hearing aid to be transferred of the user from the fitting parameters of the first-level hearing aids, and marking the fitting parameters as。
By the model: Generating unified hidden variables 。
In a specific embodiment of the invention, the method for generating the fitting parameters of the converted hearing aid comprises the steps of matching the characteristics of the converted hearing aid from the characteristics of each hearing aid, and marking the characteristics of the converted hearing aid asWherein。
Based on unified hidden variablesThe characteristics of the hearing aid after conversion are recorded asBy means ofNetwork: Generate the first Fitting parameters of a hearing aid after a change of the individual hearing aids。
The invention adopts the deep neural network algorithm, can output the hearing-identical sensing and matching parameters of different hearing aids at one time according to the user characteristics and the hearing aid characteristics, is convenient for matching the matching parameters of the hearing aids after the conversion in the actual hearing aid matching parameter conversion process, and improves the reusability of the matching parameter conversion model.
In a specific embodiment of the invention, the judging tests the hearing aidAnd testing a hearing aidThe specific analysis method comprises extracting hearing aid hearing-aid fitting parameters from database, and recording asWherein,,A number indicating the co-listening fitting parameters of each hearing aid,Representing the total number of co-listening fitting parameters of the hearing aid.
Co-listening sensing fitting parameters from each hearing aidSelected test hearing aidIs recorded as the verification parameter ofWhereinBy testing the hearing aidIs to be tested for parameters of the fittingFrom the number of the individual hearing aid featuresMedium-matching test hearing aidCorresponding featuresFrom the characteristics of the individual hearing aidsSelecting a test hearing aidIs characterized byWherein。
Hearing aid based on testFeatures of (2)Parameters of test and allocationBy conversion ofNetwork model: Generating unified hidden variables 。
Based on unified hidden variablesHearing aid for testingFeatures of (2)By means ofNetwork model: Generating a test hearing aid Is to be tested for parameters of the fitting。
Based on the hearing-aid co-listening sensing and matching parametersHearing aid for testingIs to be tested for parameters of the fittingIf (if)Then it means testing the hearing aidTest hearing aid obtained after conversion of test parameters of (a)Is to be tested for parameters of the fittingAnd testing hearing aidsIs to be tested for parameters of the fittingBelongs to the hearing-sensing matching parameters.
If it isThen it means testing the hearing aidTest hearing aid obtained after conversion of test parameters of (a)Is to be tested for parameters of the fittingAnd testing hearing aidsIs to be tested for parameters of the fittingAnd performing fine tuning training of the hearing aid fitting parameter conversion model without the hearing aid fitting parameter.
According to the invention, according to the fitting parameters of a large number of hearing aids, the fitting parameters of the hearing aids of the patients with real hearing loss and the fitting parameters converted by the fitter, the hearing-aid hearing-sensing fitting parameters are obtained, and are used as the verification data of the fitting parameter conversion model, so that the accuracy of the fitting parameter conversion model is improved, and the requirements of users in the actual use process are met.
In a specific embodiment of the invention, the fine tuning training is implemented by the specific analysis method that hearing aids are simultaneously subjected to the fitting parametersDivided into two classes, including the fitting parameters of the first class of each tested hearing aidFeatures and characteristicsWhereinCo-listening sensing fitting parameters for each test hearing aid of the second classFeatures and characteristicsWhereinAnd (2) andProviding a training data set of fitting parameters of each test hearing aid of the first classFeatures and characteristicsAnd co-listening sensing fitting parameters of each test hearing aid of the second class。
By passing throughNetwork model and conversionNetwork model:、 obtaining the conversion and test parameters of each test hearing aid 。
Co-listening sensory fit parameters based on second class of individual test hearing aidsConversion test parameters for each test hearing aidCalculation ofAnd (3) withThe mean square error between them, noted as the loss function。
By the back propagation algorithm: Calculate gradients and update Parameters of the networkPerformingFine tuning training of a network, whereinRepresenting extraction from a databaseThe learning rate of the network is set to be,The gradient operation is represented by a gradient operation,Is extracted from the database.
The invention analyzes the simultaneous hearing and sensing fitting parameters of each hearing aid and the fitting data converted by the actual hearing aid, performs fine tuning training on the fitting parameter conversion model, refines the accuracy of user hearing and sensing copying, ensures that the fitting parameter conversion model has self-supervision learning capability, and improves the stability of the fitting parameter conversion model.
Embodiment two:
Referring to FIG. 2, a schematic diagram of a fitting parameter conversion model provided by the invention is shown, and a hearing aid fitting parameter conversion part is shown on the left side, wherein the specific flow steps are that fitting data comprises user characteristics and hearing aid characteristics, and a variation self-encoder is based on the following steps Network and method for controlling the sameNetwork, firstly realizing user characteristics passing throughGenerating hidden variables by network processing, and processing the hidden variables and the characteristics of each hearing aidNetwork processing, generating the fitting parameters of each first-level hearing aid, and selecting the characteristics and the fitting parameters of the hearing aid to be converted as conversionThe input end of the network generates unified hidden variables, and finally the unified hidden variables and the characteristics of the hearing aid after conversion are used asAnd generating the fitting parameters of the hearing aid after conversion at the input end of the network.
The structure of the experimental parameter conversion model is schematically shown, the right side is a model verification and optimization part, based on the hearing-aid sensing and matching parameters, matching parameters and characteristics of the hearing aid a are selected as conversionThe input end of the network generates unified hidden variables, and then the unified hidden variables and the characteristics of the hearing aid b are selected from the database to be used as the characteristics of the hearing aid bAnd generating the fitting parameters of the converted hearing aid b at the input end of the network, judging whether the fitting parameters of the converted hearing aid b and the fitting parameters of the hearing aid a belong to the hearing-simultaneously sensing fitting parameters, and if not, executing fine tuning training.
The foregoing is merely illustrative and explanatory of the principles of this invention, as various modifications and additions may be made to the specific embodiments described, or similar arrangements may be substituted by those skilled in the art, without departing from the principles of this invention or beyond the scope of this invention as defined in the claims.
Claims (4)
1. A method for converting a plurality of hearing aid fitting parameters based on clinical hearing supervision, the method comprising:
Firstly, constructing user characteristics and hearing aid characteristics, namely acquiring basic information and hearing information of a user, constructing user characteristics, acquiring information of each hearing aid and constructing the characteristics of each hearing aid;
Step two, generating the fitting parameters of each first-level hearing aid based on the user characteristics by Network, generating hidden variable, inputting hidden variable and each hearing aid characteristicGenerating the fitting parameters of each first-level hearing aid in a network;
the hidden variable is generated by the specific analysis method that:
Based on Network, willThe potential spatial dimension of the network is set to 128;
based on the user characteristics, by model: characterizing the user As a result ofInput data of the network, throughEncoding process of network, outputting hidden variable, and recording asWhereinRepresentation ofA network processing process;
the specific analysis method for generating the fitting parameters of each first-level hearing aid comprises the following steps:
Based on hidden variables Features of the respective hearing aidsBy the model: Sequentially adding hidden variables And the firstFeatures of individual hearing aidsAs a result ofInput end data of the network, generating a first levelFitting parameters of individual hearing aidsWhereinRepresentation ofThe processing procedure of the network is that,Representing input data connection symbols;
similarly, the hidden variable and the characteristics of each hearing aid are sequentially taken as Input end data of the network is used for obtaining the verification parameters of each first-level hearing aid;
step three, generating unified hidden variables, namely matching the characteristics of the hearing aid to be converted and the characteristics of the hearing aid after conversion from the characteristics of each hearing aid, and selecting the first-level hearing aid fitting parameters corresponding to the characteristics of the hearing aid to be converted and the characteristics of the hearing aid before conversion as conversion The input of the network generates unified hidden variables;
The conversion is performed The network comprises the following specific analysis methods:
Based on Network model: By inputting hidden variables and first level Features of individual hearing aids, viaAfter the decoding processing of the network, obtain the firstFitting parameters of individual hearing aids, willReverse reconfiguration of network through conversionNetwork model: Will be at the first Fitting parameters and characteristics of individual hearing aids as steeringInput data of the network, carrying out recovery of hidden variables, whereinRepresenting the reduced hidden variable, whereinIndicating steeringA network processing process;
The specific analysis method for generating the unified hidden variable comprises the following steps:
Matching the characteristics of the hearing aid to be converted from the characteristics of each hearing aid, and marking the characteristics of the hearing aid to be converted as the characteristics of the user to be converted WhereinMatching the fitting parameters of the first-level hearing aid corresponding to the characteristics of the hearing aid to be transferred of the user from the fitting parameters of the first-level hearing aids, and marking the fitting parameters as;
By the model: Generating unified hidden variables ;
Generating the test and matching parameters of the converted hearing aid, namely taking the characteristics of the converted hearing aid and the unified hidden variables as based on the unified hidden variablesThe input of the network generates the fitting parameters of the hearing aid after conversion;
the specific analysis method for generating the fitting parameters of the converted hearing aid comprises the following steps:
matching the characteristics of the converted hearing aid from the characteristics of each hearing aid, and marking the characteristics of the converted hearing aid as Wherein;
Based on unified hidden variablesThe characteristics of the hearing aid after conversion are recorded asBy means ofNetwork: Generate the first Fitting parameters of a hearing aid after a change of the individual hearing aids;
Step five, verifying a hearing aid fitting parameter conversion model, namely selecting a test hearing aid from a databaseIs to test the hearing aidAfter the characteristics and the fitting parameters of the hearing aid are processed by a hearing aid fitting parameter conversion model, a test hearing aid is obtainedIs used for judging and testing the hearing aidAnd testing a hearing aidWhether the test matching parameters belong to the hearing test matching parameters or not is judged, and fine tuning training is implemented;
the judging and testing hearing aid And testing a hearing aidWhether the test matching parameters belong to the hearing test matching parameters or not is determined by the specific analysis method:
extracting hearing aid co-listening sensing and matching parameters from a database and marking the hearing aid co-listening sensing and matching parameters as Wherein,,A number indicating the co-listening fitting parameters of each hearing aid,Representing the total number of co-listening sensory fit parameters of the hearing aid;
Co-listening sensing fitting parameters from each hearing aid Selected test hearing aidIs recorded as the verification parameter ofWhereinBy testing the hearing aidIs to be tested for parameters of the fittingFrom the number of the individual hearing aid featuresMedium-matching test hearing aidCorresponding featuresFrom the characteristics of the individual hearing aidsSelecting a test hearing aidIs characterized byWherein;
Hearing aid based on testFeatures of (2)Parameters of test and allocationBy conversion ofNetwork model: Generating unified hidden variables ;
Based on unified hidden variablesHearing aid for testingFeatures of (2)By means ofNetwork model: Generating a test hearing aid Is to be tested for parameters of the fitting;
Based on the hearing-aid co-listening sensing and matching parametersHearing aid for testingIs to be tested for parameters of the fittingIf (if)Then it means testing the hearing aidTest hearing aid obtained after conversion of test parameters of (a)Is to be tested for parameters of the fittingAnd testing hearing aidsIs to be tested for parameters of the fittingBelongs to the hearing-sensing matching parameters;
If it is Then it means testing the hearing aidTest hearing aid obtained after conversion of test parameters of (a)Is to be tested for parameters of the fittingAnd testing hearing aidsIs to be tested for parameters of the fittingAnd performing fine tuning training of the hearing aid fitting parameter conversion model without the hearing aid fitting parameter.
2. The method for converting a plurality of hearing aid fitting parameters based on clinical hearing supervision according to claim 1, wherein the method for constructing user characteristics comprises the following specific analysis methods:
based on basic information input by a user in advance, extracting the age and sex of the user to obtain the basic information of the user;
extracting a hearing threshold value, a hearing loss type and a hearing loss level of a user based on an audiogram input by the user to obtain hearing information of the user;
Combining basic information and hearing information of a user to construct user characteristics, which are recorded as 。
3. A method for converting a plurality of hearing aid fitting parameters under clinical hearing supervision according to claim 2, wherein the method for constructing the characteristics of each hearing aid is as follows:
extracting the type and the chip type of each hearing aid from the database, combining the type and the chip type of each hearing aid to obtain information of each hearing aid, and constructing the characteristics of each hearing aid based on the information of each hearing aid, and marking as Wherein,The number of each hearing aid is indicated,Representing the total number of hearing aids;
types of hearing aids include behind-the-ear hearing aids, in-the-ear hearing aids, canal hearing aids, and bone conduction hearing aids;
the chip types of each hearing aid comprise a high-end signal processing chip, a medium-end signal processing chip, a low-end signal processing chip and an artificial intelligent chip.
4. A method for converting a plurality of hearing aid fitting parameters under clinical hearing supervision according to claim 1, wherein the performing of fine tuning training comprises the following specific analysis methods:
co-listening sensing and fitting parameters of each hearing aid Divided into two classes, including the fitting parameters of the first class of each tested hearing aidFeatures and characteristicsWhereinCo-listening sensing fitting parameters for each test hearing aid of the second classFeatures and characteristicsWhereinAnd (2) andProviding a training data set of fitting parameters of each test hearing aid of the first classFeatures and characteristicsAnd co-listening sensing fitting parameters of each test hearing aid of the second class;
By passing throughNetwork model and conversionNetwork model:、 obtaining the conversion and test parameters of each test hearing aid ;
Co-listening sensory fit parameters based on second class of individual test hearing aidsConversion test parameters for each test hearing aidCalculation ofAnd (3) withThe mean square error between them, noted as the loss function;
By the back propagation algorithm: Calculate gradients and update Parameters of the networkPerformingFine tuning training of a network, whereinRepresenting extraction from a databaseThe learning rate of the network is set to be,The gradient operation is represented by a gradient operation,Is extracted from the database.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202510019428.9A CN119421097B (en) | 2025-01-07 | 2025-01-07 | Conversion method of multiple hearing aid fitting parameters based on clinical auditory sense supervision |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202510019428.9A CN119421097B (en) | 2025-01-07 | 2025-01-07 | Conversion method of multiple hearing aid fitting parameters based on clinical auditory sense supervision |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN119421097A CN119421097A (en) | 2025-02-11 |
| CN119421097B true CN119421097B (en) | 2025-03-28 |
Family
ID=94471414
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202510019428.9A Active CN119421097B (en) | 2025-01-07 | 2025-01-07 | Conversion method of multiple hearing aid fitting parameters based on clinical auditory sense supervision |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN119421097B (en) |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116614757A (en) * | 2023-07-18 | 2023-08-18 | 江西斐耳科技有限公司 | Hearing aid fitting method and system based on deep learning |
Family Cites Families (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6990447B2 (en) * | 2001-11-15 | 2006-01-24 | Microsoft Corportion | Method and apparatus for denoising and deverberation using variational inference and strong speech models |
| EP3148213B1 (en) * | 2015-09-25 | 2018-09-12 | Starkey Laboratories, Inc. | Dynamic relative transfer function estimation using structured sparse bayesian learning |
| CN106714062B (en) * | 2016-11-30 | 2020-02-18 | 天津大学 | An intelligent fitting method for digital hearing aids based on BP artificial neural network |
| CN109151692B (en) * | 2018-07-13 | 2020-09-01 | 南京工程学院 | Hearing aid self-fitting method based on deep learning network |
| DE102019200956A1 (en) * | 2019-01-25 | 2020-07-30 | Sonova Ag | Signal processing device, system and method for processing audio signals |
| DK3709115T3 (en) * | 2019-03-13 | 2023-04-24 | Oticon As | HEARING DEVICE OR SYSTEM COMPRISING A USER IDENTIFICATION DEVICE |
| EP3836570B1 (en) * | 2019-12-12 | 2025-07-16 | Oticon A/s | Signal processing in a hearing device |
| CN112383870B (en) * | 2020-10-29 | 2022-03-18 | 惠州市锦好医疗科技股份有限公司 | Adaptive hearing parameter fitting method and device |
| EP4021017A1 (en) * | 2020-12-28 | 2022-06-29 | Oticon A/s | A hearing aid comprising a feedback control system |
| EP4274259A1 (en) * | 2022-05-02 | 2023-11-08 | Oticon A/s | A hearing aid comprising a signal processing network conditioned on auxiliary parameters |
| WO2024238643A1 (en) * | 2023-05-15 | 2024-11-21 | Dolby Laboratories Licensing Corporation | Audio processing using hearing loss data |
-
2025
- 2025-01-07 CN CN202510019428.9A patent/CN119421097B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116614757A (en) * | 2023-07-18 | 2023-08-18 | 江西斐耳科技有限公司 | Hearing aid fitting method and system based on deep learning |
Also Published As
| Publication number | Publication date |
|---|---|
| CN119421097A (en) | 2025-02-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN101924977B (en) | Method for adjusting signal processing parameters of a first hearing aid and a second hearing aid and hearing aid comprising a signal processor | |
| Moodie et al. | Procedure for predicting real-ear hearing aid performance in young children | |
| JP4718559B2 (en) | Method and apparatus for individualizing HRTFs by modeling | |
| CN112995876A (en) | Signal processing in a hearing device | |
| JP2017526236A (en) | Hearing aid system parameter optimization method and hearing aid system | |
| US20200389743A1 (en) | Method for configuring a hearing-assistance device with a hearing profile | |
| US20240267682A1 (en) | Method of operating a hearing aid system and a hearing aid system | |
| WO2021030584A1 (en) | Systems, devices and methods for fitting hearing assistance devices | |
| DK2172062T3 (en) | A method of adapting a hearing aid by means of a perceptual model | |
| Salehi et al. | Learning-based reference-free speech quality measures for hearing aid applications | |
| US8755533B2 (en) | Automatic performance optimization for perceptual devices | |
| EP1830602A1 (en) | A method of obtaining settings of a hearing instrument, and a hearing instrument | |
| CN119421097B (en) | Conversion method of multiple hearing aid fitting parameters based on clinical auditory sense supervision | |
| EP3085109A1 (en) | Method and apparatus for fitting a hearing device | |
| EP3210392B1 (en) | Method of operating a hearing aid system and a hearing aid system | |
| Sokolova | Applications of Open Source Software for Hearing Aid Amplification and Hearing Loss Simulation | |
| Bramsløw | An objective estimate of the perceived quality of reproduced sound in normal and impaired hearing | |
| Croghan et al. | Chosen listening levels for music with and without the use of hearing aids | |
| EP4387273A1 (en) | Fitting system, and method of fitting a hearing device | |
| CN117349984B (en) | Design method, design system, storage medium and server for wireless earphone | |
| Gao et al. | Next-generation hearing prosthetics | |
| Brennan et al. | The situational hearing aid response profile: An update | |
| TWM645529U (en) | Cloud hearing aid management system | |
| TW202437131A (en) | Cloud hearing aid management system and method | |
| Fierro | Adaptive loudness compensation in audio reproduction |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |