JP7276433B2 - FITTING ASSIST DEVICE, FITTING ASSIST METHOD, AND PROGRAM - Google Patents

Description

The present invention relates to a fitting support device, a fitting support method, and further to a program for realizing these.

A wearable device that is worn on a living body needs to be adjusted (fitted) to match the living body. For example, in the case of hearing aids, technicians perform fitting in consideration of the subject's hearing ability, ear structure, living environment, and the like.

However, depending on the proficiency (skill level) of the technician, there are cases where appropriate fitting is not performed for the subject. Therefore, a system for assisting fitting has been proposed.

As a related technique, Patent Literature 1 discloses a fitting device for automatically fitting a hearing aid. According to the fitting device disclosed in Patent Document 1, the subject listens to environmental sounds created in advance, evaluates how the subject hears the environmental sounds, and adjusts the hearing aid until the subject is satisfied. It is a device that performs fitting by repeating the adjustment of

Japanese Patent Application Laid-Open No. 2001-008295

However, the fitting device disclosed in Patent Document 1 can perform fitting so that the noise contained in the environmental sound is not perceived as annoying, but it is difficult to perform the fitting performed by highly skilled technicians. difficult.

In addition, when performing fitting on a wearing device, since there are various target persons, it is necessary to consider the characteristics of each target person. Therefore, it is difficult for technicians with low proficiency to perform appropriate fitting on the subject.

An example of an object of the present invention is to provide a fitting assistance device, a fitting assistance method, and a program for improving fitting accuracy.

In order to achieve the above object, the fitting assistance device in one aspect of the present invention includes:
acquisition means for acquiring test information representing results of tests performed on a subject and survey information representing results of surveys related to the subject;
estimating, in the fitting of the parameter data used to adapt the device to the subject, by inputting the obtained examination information and the survey information, and estimating a conformance index representing the degree of conformity of the parameter data and its distribution; means and
characterized by having

Further, in order to achieve the above object, the fitting assistance method in one aspect of the present invention includes:
(a) obtaining test information representing results of tests performed on a subject and survey information representing results of surveys related to the subject;
(b) inputting the obtained inspection information and the survey information in the fitting of the parameter data used to adapt the device to the subject, and estimating a conformance index representing the degree of conformity of the parameter data and its distribution; It is characterized by doing.

Furthermore, in order to achieve the above object, the program in one aspect of the present invention is
to the computer,
(a) obtaining test information representing results of tests performed on a subject and survey information representing results of surveys related to the subject;
(b) inputting the obtained inspection information and the survey information in the fitting of the parameter data used to adapt the device to the subject, and estimating a conformance index representing the degree of conformity of the parameter data and its distribution; do, step and
is characterized by executing

As described above, according to the present invention, fitting accuracy can be improved.

FIG. 1 is a diagram showing an example of a fitting assistance device. FIG. 2 is a diagram showing an example of a system having a fitting assistance device. FIG. 3 is a diagram showing an example of the data structure of selection information. FIG. 4 is a diagram showing an example of the data structure of parameter data. FIG. 5 is a diagram showing a display example of a matching index of parameter data. FIG. 6 is a diagram showing a display example of a matching index of parameter data. FIG. 7 is a diagram showing an example of the data structure of examination information. FIG. 8 is a diagram showing an example of the data structure of survey information. FIG. 9 is a diagram showing a display example of the amount of information corresponding to missing information. FIG. 10 is a diagram showing an example of a system having a learning device. FIG. 11 is a diagram showing an example of the operation of the fitting assistance device. FIG. 12 is a diagram illustrating an example of the operation of the learning device; FIG. 13 is a diagram illustrating an example of a computer that implements the fitting support device.

(Embodiment)
BEST MODE FOR CARRYING OUT THE INVENTION An embodiment of the present invention will be described below with reference to FIGS.

[Device configuration]
First, using FIG. 1, the configuration of a fitting assistance device 1 according to the present embodiment will be described. FIG. 1 is a diagram showing an example of a fitting assistance device.

The fitting support device shown in FIG. 1 is a device for improving the accuracy of adjustment (fitting) for fitting a device to a subject. Further, as shown in FIG. 1 , the fitting support device 1 has an acquisition unit 2 and an estimation unit 3 .

A device may be, for example, a device that is used by being attached to a living body. Specifically, devices include hearing aids, earphones, smart glasses, head-mounted displays, smart watches, musical instruments, medical equipment, nursing care robots, and the like. However, the device may be a device that is used without being attached to the living body.

Of these, the acquisition unit 2 acquires test information representing the results of tests performed on the subject and survey information representing the results of surveys related to the subject. The estimating unit 3 receives the acquired test information and survey information in fitting of parameter data used to match the device to the subject, and estimates a matching index representing the degree of matching of the parameter data and its distribution.

The examination information is information (explanatory variables) representing the results of examinations performed on the subject. Survey information is information (explanatory variables) representing the results of surveys related to subjects. Parameter data (objective variable) is data (setting values) used for adjusting the hardware and software of the target device.

The fit index is an index that represents the degree of fit of parameter data in fitting. As a matching index, for example, a probability representing the degree of likelihood of parameter data can be considered. Alternatively, the number that satisfies a constraint condition such as a safety standard for fitting defined based on domain knowledge may be used as the matching index.

The estimating unit 3 receives a plurality of test information acquired in the past, a plurality of survey information acquired in the past, and parameter data set in the device in the past, and generates parameter data by machine learning. It has a learning model that is used to estimate the index and its distribution.

Machine learning can be, for example, supervised learning, semi-supervised learning, and the like. Specifically, machine learning includes regression (regression techniques such as Bayesian linear regression), multi-class classification (algorithms such as decision trees), and the like.

As described above, in the present embodiment, the fitting support device 1 can estimate the parameter data match index, and therefore can present the estimated parameter data match index and its distribution to the technician. Therefore, the technician can perform appropriate fitting on the subject.

Specifically, the fitting support device 1 outputs a probability distribution to the output device as the distribution of the matching index of the parameter data, so that even a technician with a low skill level can refer to the output probability distribution A suitable fitting can be performed for Therefore, it is possible to improve the skill of a technician whose skill level is low, so that the accuracy of fitting can be improved.

[System configuration]
Next, with reference to FIG. 2, the configuration of the fitting assistance device 1 according to the present embodiment will be described more specifically. FIG. 2 is a diagram showing an example of a system having the fitting support device 1. As shown in FIG.

As shown in FIG. 2, the system having the fitting assistance device 1 in this embodiment has an input device 21 and an output device 22 in addition to the fitting assistance device 1 . However, the system is not limited to the configuration shown in FIG.

The fitting support device 1 (1) estimates the matching index of the parameter data and its distribution. In addition, the fitting support device 1 (2) estimates the content of the survey to be performed on the subject in order to determine the parameter data.

(1) Estimation of parameter data compatibility index and its distribution In (1), instead of presenting only one candidate parameter data to the technician as in the past, even the compatibility index and its distribution are estimated. By doing so, we present the entire distribution and several promising candidate parameters.

When only one candidate parameter data is presented to the technician as in the conventional method, the subject is asked to evaluate whether the equipment set using the candidate parameter data is suitable for him/herself. Then, when the subject evaluates that it is suitable, the fitting is terminated. However, in actual fitting, it is difficult to match the equipment to the subject in one fitting.

If the subject is evaluated as non-conforming, the technician must modify the parameters. However, a technician with a low skill level does not know how to correct the parameter data and repeats trial and error, resulting in a long time required for fitting.

In other words, since the specific fitting policy is unknown, the fitting takes a long time. Furthermore, the final parameters used to calibrate the device may not actually match the subject.

Therefore, in (1), by estimating even the matching index of parameter data and its distribution, and presenting the entire distribution and multiple promising candidate parameters to the technician, the time required for the technician to perform fitting is reduced. to shorten In addition, by using candidate parameters, the parameters can be easily matched to the subject.

(2) Estimation of survey contents to be conducted on the subject to determine parameter data Explanatory variables such as information, survey information, etc. are required. However, in fitting, explanatory variables such as necessary inspection information and survey information may not be available (shortage).

Then, the determined parameter data does not actually fit the subject because of the lack of explanatory variables. Therefore, in (2), research is conducted to obtain the necessary explanatory variables, and the necessary explanatory variables are estimated based on the research results.

Specifically, in (2), the content of questions, the content of inspection, etc. (survey content) required to obtain the necessary explanatory variables are obtained and presented to the technician, the subject, or both. Subsequently, a response to the survey content, ie, missing examination information, or survey information, or both (explanatory variables) is obtained. After that, the technician causes the fitting support device 1 to estimate the parameter data again using the explanatory variables.

In this way, according to (2), since the research contents can be presented, a specific fitting policy can be clarified, and the time required for the fitting can be shortened by the technician. In addition, since necessary explanatory variables can be arranged as much as possible, the accuracy of estimating parameters suitable for the subject can be improved.

The fitting support device 1 has a selection unit 24 and an output information generation unit 25 in addition to the acquisition unit 2 and the estimation unit 3 . Furthermore, the estimation unit 3 has a learning model 26 . The selection unit 24 has a learning model 27 and a learning model 28 . However, the learning models 26 , 27 and 28 may be provided inside the fitting assistance device 1 or may be provided outside the fitting assistance device 1 .

The input device 21 is a device used to input examination information and survey information to the fitting support device 1 . Specifically, the input device 21 first acquires the subject's test information and survey information from an information processing device or a storage device provided in equipment dealers, manufacturers, related facilities, and the like. Subsequently, the input device 21 transmits the obtained examination information and survey information to the obtaining unit 2 of the fitting support device 1 using wired or wireless communication.

In addition, if the input device 21 finds missing information in the examination information, investigation information, or both of them used for fitting, and if the missing information can be acquired after that, the acquired information is used for fitting assistance. Used for additional input to the device 1 .

Missing information refers to explanatory variables (variables of unobserved data) that have not yet been acquired when the explanatory variables required to obtain parameter data (objective variable) are known.

Note that the input device 21 is, for example, an information processing device such as a personal computer, a mobile computer, a smart phone, or a tablet. Alternatively, a plurality of input devices 21 may be prepared and examination information and investigation information may be input from separate input devices 21 .

The output device 22 acquires output information converted into a format that can be output by the output information generation unit 25, and outputs images and sounds generated based on the output information. The output device 22 is, for example, an image display device using liquid crystal, organic EL (Electro Luminescence), or CRT (Cathode Ray Tube). Furthermore, the image display device may include an audio output device such as a speaker. Note that the output device 22 may be a printing device such as a printer.

The fitting assistance device 1 will be described in detail.
The acquisition unit 2 acquires examination information and investigation information from the input device 21 in the operation phase. Specifically, the acquisition unit 2 receives test information representing the results of tests performed on the subject and survey information representing the results of surveys related to the subject, which are transmitted from the input device 21. .

In the operation phase, the estimating unit 3 receives the obtained inspection information and survey information as input, and estimates a conformity index representing the degree of conformity of the parameter data and its distribution. Furthermore, if there is missing information in the test information, survey information, or both of them used for fitting, the estimating unit 3 acquires the missing information, adds the acquired information, and re- , to estimate the fit index of the parameter data and its distribution.

Specifically, the estimation unit 3 first acquires examination information and investigation information (explanatory variables). Subsequently, the estimating unit 3 inputs the obtained inspection information and investigation information (explanatory variables) to the learning model 26, and estimates a conformity index representing the degree of conformity of the parameter data and its distribution.

For example, when all the explanatory variables to be input to the learning model 26 are available, the input can be expressed as Equation 1.

[Number 1]

Then, when all the explanatory variables to be input are available, the distribution of the matching index of the parameter data (objective variable) can be expressed as Equation 2 using, for example, probability distribution.

また、学習モデル２６へ入力する説明変数が不足している場合、入力は数３のように表すことができる。数３の例では、不足している情報を未観測データＺ_３、Ｚ_４として表している。[Number 2]

Also, when explanatory variables to be input to the learning model 26 are insufficient, the input can be expressed as in Equation 3. In the example of Expression 3, the missing information is represented as unobserved data Z ₃ and Z ₄ .

[Number 3]

Moreover, when there are unobserved data Z ₃ and Z ₄ , the probability distribution of the parameter data (objective variable) can be expressed as Equation 4.

[Number 4]

However, although the unobserved data Z ₃ and Z ₄ are not dependent on observed data other than the unobserved data Z ₃ and Z ₄ in Expression 4, they may be dependent.

In the learning phase, the learning model 26 includes test information (explanatory variables) of a plurality of users acquired in the past, research information (explanatory variables) of a plurality of users acquired in the past, and It is a learning model generated by machine learning with parameter data as input. Also, the learning model 26 inputs explanatory variables and outputs matching indices and their distributions for each objective variable of the parameter data. Details of the learning model 26 will be described later.

Although the learning model 26 is provided inside the fitting assistance device 1 in FIG. good. In that case, the estimation unit 3 is configured to be able to communicate with the information processing device or the storage device.

In the operation phase, if there is missing information in the inspection information, the investigation information, or both of them used for fitting, the selection unit 24 associates the missing information with the missing information to make up for the missing information. Select the question information

Specifically, the selection unit 24 first detects whether or not there is missing information in the examination information, investigation information, or both of them (explanatory variables) used for fitting.

Subsequently, if there is missing information (explanatory variables), the selection unit 24 selects information used to obtain the missing information (for example, a binary value indicating whether information has been acquired for each explanatory variable). vectors as elements) are input to the learning model 27, and the probability distribution of the missing information is estimated. The probability distribution of missing information (appearance frequency in learning data) can be expressed as in Equation 5.

[Number 5]

In the learning phase, the learning model 27 is generated by machine learning, with input of test information (explanatory variables) of multiple users acquired in the past and survey information (explanatory variables) of multiple users acquired in the past. It is a learning model that In addition, the learning model 27 stores information used to obtain missing information for each inspection information and investigation information (explanatory variable) (for example, a vector whose elements are binary values representing whether or not information has been acquired). Input and output the probability distribution of each unacquired explanatory variable. Details of the learning model 27 will be described later.

In addition, if there is missing information, the selection unit 24 selects information used to obtain the missing information (for example, a vector whose elements are binary values representing whether or not information has been acquired for each explanatory variable). is input to the learning model 28 to estimate the joint distribution of the parameter data and the missing information. The joint distribution can be expressed as in Equation 6.

[Number 6]

In the learning phase, the learning model 28 includes previously acquired examination information (explanatory variables) of multiple users, previously acquired survey information (explanatory variables) of multiple users, and previously set devices. It is a learning model generated by machine learning with parameter data (objective variable) as input. Also, the learning model 28 inputs examination information, investigation information (explanatory variables), and parameter data (objective variables), and outputs a joint distribution of the explanatory variables and the objective variables. Details of the learning model 28 will be described later.

Subsequently, the selection unit 24 calculates the amount of information (importance) for each missing piece of information using the estimated probability distribution and the joint distribution. The amount of information can be expressed as Equation 7.

[Number 7]

Next, the selection unit 24 calculates the amount of information representing the correlation with the parameter data (objective variable) for each missing piece of information, and based on this amount of information, selects one of the pieces of missing information. Select the above information. The amount of information can be, for example, the amount of information with conditions. The selection unit 24 preferably selects the missing information in descending order of importance indicated by the amount of information.

Subsequently, based on the selected missing information, the selection unit 24 refers to the selection information in which the identification information that identifies the explanatory variable and the question information are associated, and selects the question information. Question information is, for example, information representing the content of a question to be asked to a subject in order to acquire missing information. The selection information is stored, for example, in a storage device provided inside or outside the fitting assistance device 1 .

FIG. 3 is a diagram showing an example of the data structure of selection information. FIG. 3 shows an example of selection information when the device is a hearing aid. Here, the hearing aid, for example, collects sound using a sound collecting unit (microphone), amplifies and processes the collected sound using a processing unit, and amplifies and processes using an output unit (receiver). It is a device that outputs sound.

In the selection information of FIG. 3, identification information for identifying test information and survey information is stored in "item", and question information representing question content used to acquire test information and survey information is stored in "question content". is stored.

The output information generation unit 25 generates output information used for outputting the matching index of the parameter data estimated by the estimation unit 3 to the output device 22 . The output information generator 25 then outputs the generated output information to the output device 22 .

FIG. 4 is a diagram showing an example of the data structure of parameter data. Also, FIG. 4 shows parameter data when the device is a hearing aid. In the example of FIG. 4, the parameter data is, for example, data used for adjusting one or more of the sound collecting unit, the processing unit, the output unit, or each of these units provided in the hearing aid.

In addition, as shown in FIG. 4, each of the data (objective variables) included in the parameter data is frequency Characteristic value, noise suppression strength, howling suppression strength, directivity type, impact noise suppression degree, etc. can be considered.

Furthermore, in the example of FIG. 4, the parameter data includes "item" information representing the objective variable item of the parameter data, "parameter" information representing the estimated parameter data (objective variable), and data type. The data type is associated with the information of the "data type".

It is conceivable to display the distribution of the matching index of the parameter data as shown in FIG. 5, for example. FIG. 5 is a diagram showing a display example of the distribution of matching indices of parameter data.

In the case of the frequency characteristic value in the "item" of FIG. 4, 50 [dB], the objective variable is the output level 30 [dB], 40 [dB ] . . . Therefore, to assist in determining these objective variables, the estimated probability distribution for each objective variable is displayed. By displaying the probability distribution for each objective variable to be displayed in this way, it is possible to present a plurality of objective variable candidates to the technician.

Note that a display as shown in FIG. 6 may be used. FIG. 6 is a diagram showing a display example of the distribution of matching indices of parameter data. Also, the probability distribution for each objective variable may be displayed using a histogram. Furthermore, the magnitude of probability may be displayed using a heat map.

The output information generation unit 25 also generates output information used to output the question content represented by the question information selected by the selection unit 24 to the output device 22 . The output information generator 25 then outputs the generated output information to the output device 22 .

For example, if the device is a hearing aid, it is assumed that the test information and survey information have explanatory variables as shown in FIGS. 7 and 8. FIG. FIG. 7 is a diagram showing an example of the data structure of examination information. FIG. 8 is a diagram showing an example of the data structure of survey information.

In the example of FIG. 7, if the device is a hearing aid, the test information includes at least one or more of air conduction audiogram, bone conduction audiogram, discomfort threshold, and speech intelligibility. In the example of FIG. 7, the test information includes "item" information representing items of the hearing test, information of "hearing test result" representing the hearing test result, and "data type" representing the type of data. Information is associated and stored in a storage device.

In the example of FIG. 8, the survey information includes, for example, at least the subject's age, sex, occupation, residence, family composition, medical history, treatment history, device usage history, symptoms, and information on physical characteristics (e.g., height, weight, etc.). , ear acoustics, etc.). Note that ear acoustics is information representing the acoustic characteristics of the ear.

In addition, the survey information may include information representing the subject's living environment sounds, information representing preferences, and adjustment information. A living environment sound is information representing a sound heard in a subject's daily life. The preference is information representing the subject's preference for sound. The adjustment information is information representing the date and time of adjustment of the device, the place of adjustment, the identification number of the adjuster, and the like. Furthermore, the attribute information may include information representing the type of hearing aid.

In the example shown in FIG. 8, the adjustment information includes "item" information representing survey items (attributes, background, etc.) related to the subject, and "survey content" information representing the survey content of the subject. , and "data type" information representing the type of data, and stored in the storage device.

When such explanatory variables are arranged, for example, if the missing information is "speech intelligibility" and "age", the question content "speech intelligibility" shown in FIG. Please input the result of the examination.” and “Please input your age.”

Furthermore, the output information generation unit 25 generates output information used for outputting the amount of information calculated by the selection unit 24 to the output device 22 . The output information generator 25 then outputs the generated output information to the output device 22 .

FIG. 9 is a diagram showing a display example of the amount of information corresponding to missing information. For example, it is conceivable to display a graph as shown in FIG. 9 to show the amount of information for missing information.

Next, generation of a learning model will be described.
FIG. 10 is a diagram showing an example of a system having a learning device. A learning device 31 shown in FIG. 10 is a device that generates learning models 26, 27, and 28 using machine learning.

A system having a learning device 31 shown in FIG. 10 has a storage device 32 in addition to the learning device 31 . The learning device 31 also has an acquisition unit 33 , a classification unit 34 , a classification unit 35 , and a generation unit 36 .

The storage device 32 stores explanatory variables (examination information, investigation information) and objective variables (parameter data) acquired in the past.

A plurality of test information acquired in the past is information representing results of tests performed by users of a plurality of devices in the past for each user. A plurality of pieces of adjustment information acquired in the past are pieces of adjustment information acquired by users of a plurality of devices in the past for each user.

The parameter data acquired in the past is the parameter data used for device adjustment in past fittings performed by technicians with a high skill level on users of a plurality of devices. Further, the parameter data acquired in the past is parameter data used for adjustment of the devices in the past, when fittings were performed for users of a plurality of devices using the fitting support apparatus 1 .

A learning device will be described.
The acquisition unit 33 acquires learning data acquired in the past. Specifically, the acquisition unit 33 acquires learning data such as a plurality of pieces of examination information, adjustment information, and parameter data acquired in the past from the storage device 32 , and transmits the acquired learning data to the classification unit 34 .

The classification unit 34 classifies the received learning data. Specifically, the classification unit 34 first compares a score (satisfaction level) indicating whether or not the user was satisfied with the fitting in the past with a preset threshold value. Subsequently, when the score is equal to or greater than the threshold, the classification unit 34 classifies the learning data associated with the score equal to or greater than the threshold.

In this way, the classification unit 34 can make the learning models 26, 27, and 28 learn using only learning data with a high degree of satisfaction, so the parameter data estimated using the learning models 26, 27, and 28 are , user satisfaction increases.

The classification unit 35 further classifies the learning data classified by the classification unit 34 . Specifically, the classification unit 35 further performs clustering processing on the learning data classified by the classification unit 34 .

In clustering processing, high-dimensional information is transformed into low-dimensional information. The reason for this is that the explanatory variables that can be represented by sentences and time-series signals become variables with enormous numerical values and categorical dimensions, and if they are learned as they are, the amount of computational resources increases. Therefore, evaluation of the presence or absence of each word, TF-IDF, etc. are applied to the character string data, converted to a numerical vector, and this vectorized data is processed, for example, by the k-mean method, deep learning, etc. can be used to transform high-dimensional information into low-dimensional information.

As an example, first, high-dimensional information (eg, character strings, time-series signals, etc.) is input to a deep neural network and converted into vector values. Subsequently, the transformed vector values are transformed into low-dimensional information (for example, labels, etc.) using the k-nearest neighbor method.

In addition to classifying with supervised learning using labels created from domain knowledge, unsupervised learning can be performed by using a clustering method such as the k-means method even if labels are not assigned in advance. You can make labels.

In this way, in the classification unit 34, the learning models 26, 27, and 28 can be trained using only learning data with a high degree of satisfaction. increases user satisfaction. Furthermore, since learning data can be reduced, it is possible to reduce the amount of computational resources, such as shortening the time required for learning and reducing the amount of memory used.

Moreover, in the classification unit 35, by performing clustering processing on the learning data classified by the classification unit 34, it is possible to further reduce the amount of computational resources in the case of learning.

The generation unit 36 performs machine learning using the learning data classified by the classification unit 35 to generate learning models 26 , 27 and 28 , and stores the generated learning models 26 , 27 and 28 in the estimation unit 3 . Alternatively, the generation unit 36 may store the learning models 26, 27, and 28 in a system having the learning device 31, a system having the fitting support device 1, or other storage devices.

Note that machine learning can be, for example, supervised learning, semi-supervised learning, or the like. For example, it is conceivable to use learning such as regression (regression methods such as least squares method, Bayesian linear regression, and random forest), multi-class classification (algorithm such as decision tree), and the like. In addition, it is not limited to the machine learning mentioned above, You may use other than the machine learning mentioned above.

However, the input of the generation unit 36 may be the learning data classified by the classification unit 35, the learning data classified by the classification unit 34, or the unclassified learning data. may be used.

Also, the learning data used to generate the learning models 26, 27, and 28 may use inspection information, adjustment information, and parameter data that have been determined to be conforming in the past.

In this manner, the learning device 31 can generate the learning models 26, 27, and 28 that take into consideration not only the inspection information acquired in the past, but also the adjustment information and parameter data acquired in the past. In particular, learning models 26, 27, and 28 can be generated using the results of fitting performed by highly skilled technicians as inputs.

[Device operation]
Next, the operation of the fitting assistance device according to the embodiment of the present invention will be explained using FIG. FIG. 11 is a diagram showing an example of the operation of the fitting assistance device. In the following description, reference will be made to FIGS. 2 to 9 as appropriate. Further, in the present embodiment, the fitting assistance method is carried out by operating the fitting assistance device. Therefore, the description of the fitting support method in the present embodiment is replaced with the following description of the operation of the fitting support device.

Also, the operation of the learning device according to the embodiment of the present invention will be described with reference to FIG. FIG. 12 is a diagram illustrating an example of the operation of the learning device; In the following description, FIG. 10 will be referred to as appropriate. Further, in the present embodiment, the learning method is carried out by operating the learning device. Therefore, the description of the learning method in this embodiment is replaced with the description of the operation of the learning device below.

The operation of the fitting support device will be described.
As shown in FIG. 11, first, in the operation phase, the acquisition unit 2 acquires examination information and research information (explanatory variables) of a subject from the input device 21 (step A1). Specifically, in step A1, the acquisition unit 2 acquires test information indicating the results of tests performed on the subject and survey information indicating the results of surveys related to the subject, which are transmitted from the input device 21. and receive.

Next, in the operation phase, the estimating unit 3 receives the obtained inspection information and investigation information as input and estimates a matching index representing the matching level of the parameter data and its distribution (step A2). Specifically, in step A2, the estimation unit 3 first acquires examination information and research information (explanatory variables). Subsequently, in step A2, the estimating unit 3 inputs the obtained inspection information and survey information (explanatory variables) to the learning model 26, and estimates a conformity index representing the degree of conformity of the parameter data and its distribution.

Next, in the operation phase, if there is missing information in the examination information, survey information, or both of them used for fitting, the selection unit 24 selects question information to be used to compensate for the missing information. (Step A3). Specifically, in step A3, the selection unit 24 first detects whether or not there is missing information in the examination information, investigation information, or both (explanatory variables) used for fitting.

Subsequently, in step A3, if there is missing information (explanatory variables), the selection unit 24 selects information used to obtain the missing information (for example, whether or not information has been acquired for each explanatory variable). A vector whose elements are binary values representing ) is input to the learning model 27, and the probability distribution of the missing information is estimated. Further, in step A3, if there is missing information, the selection unit 24 selects information used to obtain the missing information (for example, a binary value representing whether information has been acquired for each explanatory variable as an element ) are input to the learning model 28 to estimate the joint distribution of the parameter data and the missing information.

Subsequently, in step A3, the selection unit 24 calculates the amount of information (importance) for each missing piece of information using the estimated probability distribution and the joint distribution. Subsequently, in step A3, the selection unit 24 calculates the amount of information representing the correlation with the parameter data (objective variable) for each missing piece of information, and based on this amount of information, Select one or more information from

Thereafter, in step A3, the selection unit 24 refers to selection information in which identification information for identifying explanatory variables and question information are associated with each other based on the selected missing information, and selects question information.

Next, the output information generation unit 25 generates output information used for outputting the matching index of the parameter data estimated by the estimation unit 3 to the output device 22 (step A4). Then, the output information generator 25 outputs the generated output information to the output device 22 (step A5).

It is conceivable that the matching index of the parameter data is displayed as shown in FIGS. 5 and 6, for example. Also, the probability distribution for each objective variable may be displayed using a histogram. Furthermore, the magnitude of probability may be displayed using a heat map, contour map, or the like.

Further, in step A4, the output information generation unit 25 generates output information used for outputting the question content represented by the question information selected by the selection unit 24 to the output device 22. FIG. Then, in step A5, the output information generating section 25 outputs the generated output information to the output device 22. FIG.

Furthermore, in step A4, the output information generation unit 25 generates output information used for outputting the amount of information calculated by the selection unit 24 to the output device 22. FIG. Then, in step A5, the output information generating section 25 outputs the generated output information to the output device 22. FIG.

In this way, the processing from steps A1 to A5 is repeated until the inspection information and investigation information (explanatory variables) necessary for determining parameter data are obtained.

If there is missing information, in step A1, the missing information acquired using the question content is added, and steps A2 to A5 are executed again.

The operation of the learning device will be described.
As shown in FIG. 12, first, the acquiring unit 33 acquires learning data acquired in the past (step B1). Specifically, in step B1, the acquisition unit 33 acquires learning data such as a plurality of test information, survey information, and parameter data acquired in the past from the storage device 32, and sends the acquired learning data to the classification unit 34. Send.

Next, the classification unit 34 classifies the received learning data (step B2). Specifically, in step B2, the classification unit 34 first compares a score (satisfaction level) indicating whether or not the user was satisfied with the fitting in the past with a preset threshold value. Subsequently, in step B2, when the score is equal to or greater than the threshold, the classification unit 34 classifies the learning data associated with the score equal to or greater than the threshold.

Next, the classification unit 35 further classifies the learning data classified in step B2 (step B3). Specifically, in step B3, the classification unit 35 further performs clustering processing on the learning data classified in step B2. In clustering processing, high-dimensional information is transformed into low-dimensional information.

Next, the generation unit 36 performs machine learning using the learning data classified by the classification unit 35 to generate the learning models 26, 27, and 28 (step B4), and the generated learning models 26, 27, and 28 are It is stored in the estimation unit 3 (step B5). Alternatively, in step B5, the generation unit 36 may store the learning models 26, 27, and 28 in the system having the learning device 31, the system having the fitting support device 1, or other storage devices.

However, the input of the generation unit 36 may be the learning data classified by the classification unit 35, the learning data classified by the classification unit 34, or the unclassified learning data. may be used.

Also, the learning data used to generate the learning models 26, 27, and 28 may use inspection information, adjustment information, and parameter data that have been determined to be conforming in the past.

[Effects of this embodiment]
As described above, according to the present embodiment, the fitting support device 1 can estimate the parameter data conformity index and its distribution, and thus can present the estimated parameter data conformity index and its distribution to the technician. Therefore, it is possible to present a plurality of candidate parameters to the technician.

Specifically, the fitting support device 1 outputs a probability distribution to the output device as the distribution of the matching index of the parameter data, so that even a technician with a low skill level can refer to the output probability distribution A suitable fitting can be performed for Therefore, it is possible to improve the skill of a technician whose skill level is low, so that the accuracy of fitting can be improved.

In addition, by estimating the matching index and its distribution of parameter data and presenting multiple candidate parameters to the technician, it is possible to present a specific fitting policy, reducing the time required for the technician to perform fitting. can be shortened. In addition, by using candidate parameters, it is possible to easily match the parameters to the subject.

In addition, since the research content can be presented, a specific fitting policy can be clarified. Therefore, the time required for the fitting by the technician can be shortened. In addition, since necessary explanatory variables can be arranged as much as possible, the accuracy of estimating parameters suitable for the subject can be improved.

[program]
The program for estimating the parameter data matching index and its distribution in the embodiment of the present invention may be a program that causes a computer to execute steps A1 to A5 shown in FIG. By installing this program in a computer and executing it, the fitting assistance device and the fitting assistance method according to the present embodiment can be realized. In this case, the processor of the computer functions as the acquiring unit 2, the estimating unit 3, the selecting unit 24, and the output information generating unit 25, and performs processing.

Also, the program for estimating the parameter data matching index and its distribution in the present embodiment may be executed by a computer system constructed by a plurality of computers. In this case, for example, each computer may function as one of the acquiring unit 2, the estimating unit 3, the selecting unit 24, and the output information generating unit 25, respectively.

Furthermore, the program for generating the learning model according to the embodiment of the present invention may be any program that causes a computer to execute steps B1 to B5 shown in FIG. By installing this program in a computer and executing it, the learning device and the learning method according to the present embodiment can be realized. In this case, the processor of the computer functions as an acquisition unit 33, a classification unit 34, a classification unit 35, and a generation unit 36 to perform processing.

Also, the program for generating the learning model in this embodiment may be executed by a computer system constructed by a plurality of computers. In this case, for example, each computer may function as one of the acquisition unit 33, the classification unit 34, the classification unit 35, and the generation unit 36, respectively.

[Physical configuration]
Here, a computer that implements the fitting support device or the learning device by executing the program according to the embodiment will be described with reference to FIG. 13 . FIG. 13 is a block diagram showing an example of a computer that implements the fitting support device or learning device according to the embodiment of the present invention.

As shown in FIG. 13, a computer 110 includes a CPU (Central Processing Unit) 111, a main memory 112, a storage device 113, an input interface 114, a display controller 115, a data reader/writer 116, and a communication interface 117. and These units are connected to each other via a bus 121 so as to be able to communicate with each other. The computer 110 may include a GPU (Graphics Processing Unit) or an FPGA (Field-Programmable Gate Array) in addition to the CPU 111 or instead of the CPU 111 .

The CPU 111 expands the programs (codes) of the present embodiment stored in the storage device 113 into the main memory 112 and executes them in a predetermined order to perform various calculations. Main memory 112 is typically a volatile storage device such as a DRAM (Dynamic Random Access Memory). Also, the program in the present embodiment is provided in a state stored in computer-readable recording medium 120 . Note that the program in this embodiment may be distributed on the Internet connected via communication interface 117 .

Further, as a specific example of the storage device 113, in addition to a hard disk drive, there is a semiconductor storage device such as a flash memory. Input interface 114 mediates data transmission between CPU 111 and input devices 118 such as a keyboard and mouse. The display controller 115 is connected to the display device 119 and controls display on the display device 119 .

Data reader/writer 116 mediates data transmission between CPU 111 and recording medium 120 , reads programs from recording medium 120 , and writes processing results in computer 110 to recording medium 120 . Communication interface 117 mediates data transmission between CPU 111 and other computers.

Specific examples of the recording medium 120 include general-purpose semiconductor storage devices such as CF (Compact Flash (registered trademark)) and SD (Secure Digital); magnetic recording media such as flexible disks; An optical recording medium such as a ROM (Compact Disk Read Only Memory) can be mentioned.

The fitting support device 1 or the learning device 31 according to the present embodiment can also be realized by using hardware corresponding to each part instead of a computer in which a program is installed. Furthermore, the fitting support device 1 or the learning device 31 may be partly implemented by a program and the rest by hardware.

[Appendix]
Further, the following additional remarks are disclosed with respect to the above embodiment. Some or all of the embodiments described above can be expressed by the following (Appendix 1) to (Appendix 18), but are not limited to the following description.

(Appendix 1)
an acquisition unit that acquires test information representing results of tests performed on a subject and survey information representing results of surveys related to the subject;
estimating, in the fitting of the parameter data used to adapt the device to the subject, by inputting the obtained inspection information and the survey information, and estimating a conformance index representing the degree of conformity of the parameter data and its distribution; Department and
A fitting support device comprising:

(Appendix 2)
The fitting assistance device according to Appendix 1,
an output information generating unit that generates output information representing the matching index and its distribution of the parameter data and outputs the output information to an output device;
A fitting support device comprising:

(Appendix 3)
The fitting assistance device according to appendix 2,
The estimating unit receives test information of a plurality of users acquired in the past, survey information of a plurality of users acquired in the past, and parameter data set in the device in the past, and generates by machine learning. A fitting support device comprising: a learning model used for estimating a matching index of said parameter data and its distribution.

(Appendix 4)
The fitting assistance device according to appendix 3,
If there is missing information in the test information, the survey information, or both used for fitting, select question information associated with the missing information to fill in the missing information. a selection unit;
The fitting support device, wherein the output information generation unit generates the output information representing the content of the question using the question information, and outputs the output information to an output device.

(Appendix 5)
The fitting assistance device according to appendix 4,
The selection unit calculates an information amount representing a correlation between the missing information and the parameter data for the missing information, and determines the missing information based on the calculated information amount. A fitting support device characterized by selecting information.

(Appendix 6)
The fitting assistance device according to appendix 4 or 5,
When the missing information is acquired using the question information, the estimating unit adds the acquired missing information and re-estimates the matching index of the parameter data and its distribution. A fitting support device characterized by having:

(Appendix 7)
(a) obtaining test information representing results of tests performed on a subject and survey information representing results of surveys related to the subject;
(b) inputting the obtained inspection information and the survey information in the fitting of the parameter data used to adapt the device to the subject, and estimating a conformance index representing the degree of conformity of the parameter data and its distribution; do, step and
A fitting support method comprising:

(Appendix 8)
The fitting support method according to appendix 7,
(c) generating output information representing the matching index and its distribution of the parameter data and outputting it to an output device;
A fitting support method comprising:

(Appendix 9)
The fitting support method according to appendix 8,
In the step (b) above, the machine receives as input the examination information of a plurality of users acquired in the past, the survey information of a plurality of users acquired in the past, and the parameter data set in the device in the past. A fitting support method characterized by using a learning model generated by learning and used for estimating a matching index and its distribution of the parameter data.

(Appendix 10)
The fitting support method according to Appendix 9,
(d) if there is missing information in the examination information, or the survey information, or both, used for fitting, questions associated with the missing information to fill in the missing information; selecting information, having the step of
The fitting support method, wherein in step (c), the output information representing the content of the question is generated using the question information and output to an output device.

(Appendix 11)
The fitting support method according to Appendix 10,
In the processing of (d), an amount of information representing the correlation between the missing information and the parameter data is calculated for the missing information, and based on the calculated amount of information, the missing A fitting support method, characterized by selecting the information that is being used.

(Appendix 12)
The fitting support method according to Appendix 10 or 11,
In the process (b), if the missing information is acquired using the question information, the acquired missing information is added, and the matching index of the parameter data and its A fitting aid method characterized by estimating a distribution.

(Appendix 13)
to the computer,
(a) obtaining test information representing results of tests performed on a subject and survey information representing results of surveys related to the subject;
(b) inputting the obtained inspection information and the survey information in the fitting of the parameter data used to adapt the device to the subject, and estimating a conformance index representing the degree of conformity of the parameter data and its distribution; do, step and
program to run.

(Appendix 14)
The program according to Appendix 13,
The program causes the computer to:
(c) A program for executing a step of generating output information representing a matching index and its distribution of the parameter data and outputting the output information to an output device.

(Appendix 15)
There is a program described in Appendix 14,
In the step (b) above, the machine receives as input the examination information of a plurality of users acquired in the past, the survey information of a plurality of users acquired in the past, and the parameter data set in the device in the past. A program characterized by using a learning model generated by learning and used for estimating a matching index of the parameter data and its distribution.

(Appendix 16)
The program according to Appendix 15,
The program causes the computer to:
(d) if there is missing information in the examination information, or the survey information, or both, used for fitting, questions associated with the missing information to fill in the missing information; further comprising instructions for performing steps for selecting information;
A program, wherein in step (c), the output information representing the content of the question is generated using the question information and output to an output device.

(Appendix 17)
The program according to Appendix 16,
In the step (d), an amount of information representing the correlation between the missing information and the parameter data is calculated for the missing information, and based on the calculated amount of information, the missing A program characterized by selecting information that is

(Appendix 18)
The program according to Appendix 16 or 17,
In step (b), when the missing information is obtained using the question information, the obtained missing information is added, and the matching index of the parameter data and its A program characterized by estimating distributions.

Although the present invention has been described with reference to the embodiments, the present invention is not limited to the above embodiments. Various changes that can be understood by those skilled in the art can be made to the configuration and details of the present invention within the scope of the present invention.

As described above, according to the present invention, fitting accuracy can be improved. INDUSTRIAL APPLICABILITY The present invention is useful in fields that require fitting for fitting a device or the like that is worn on a living body to a subject. Specifically, devices include hearing aids, earphones, smart glasses, head-mounted displays, smart watches, musical instruments, medical equipment, nursing care robots, and the like. Also, the device may be a device that is used without being attached to the living body.

1 fitting support device 2 acquisition unit 3 estimation unit 21 input device 22 output device 24 selection unit 25 output information generation unit 26, 27, 28 learning model 31 learning device 32 storage device 33 acquisition unit 34, 35 classification unit 36 generation unit 110 computer 111 CPUs
112 Main memory 113 Storage device 114 Input interface 115 Display controller 116 Data reader/writer 117 Communication interface 118 Input device 119 Display device 120 Recording medium 121 Bus

Claims (9)

acquisition means for acquiring test information representing results of tests performed on a subject and survey information representing results of surveys related to the subject;
estimating, in the fitting of the parameter data used to adapt the device to the subject, by inputting the obtained inspection information and the survey information, and estimating a conformance index representing the degree of conformity of the parameter data and its distribution; means and
If there is missing information in the inspection information used for fitting, or in the investigation information, or in both of them, the amount of information representing the correlation between the missing information and the parameter data is added to the missing information. a selection means for selecting the missing information based on the calculated amount of information;
A fitting support device comprising: The fitting assistance device according to claim 1,
output information generating means for generating output information representing the matching index and its distribution of the parameter data and outputting it to an output device;
A fitting support device comprising: The fitting assistance device according to claim 2,
The estimating means is generated by machine learning with input of test information of a plurality of users obtained in the past, survey information of a plurality of users obtained in the past, and parameter data set in the device in the past. A fitting support device comprising: a learning model used for estimating a matching index of said parameter data and its distribution. The fitting assistance device according to claim 3,
If there is missing information in the test information, the survey information, or both used for fitting, select question information associated with the missing information to fill in the missing information. to, a selection means, and
The fitting support apparatus, wherein the output information generating means generates the output information representing the content of the question using the question information, and outputs the output information to an output device. The fitting assistance device according to claim 4 ,
When the missing information is acquired using the question information, the estimating means adds the acquired missing information and re-estimates the matching index of the parameter data and its distribution. A fitting support device characterized by having: (a) obtaining test information representing results of tests performed on a subject and survey information representing results of surveys related to the subject;
(b) inputting the obtained inspection information and the survey information in the fitting of the parameter data used to adapt the device to the subject, and estimating a conformance index representing the degree of conformity of the parameter data and its distribution; do
(c) if there is missing information in the inspection information used for fitting, or in the investigation information, or both, the amount of information representing the correlation between the missing information and the parameter data is calculating for the information that is available, and selecting the missing information based on the calculated amount of information;
A fitting support method characterized by: The fitting support method according to claim 6 ,
(d) generating output information representing the matching index and its distribution of the parameter data and outputting it to an output device;
A fitting support method characterized by: The fitting support method according to claim 7 ,
In the processing of (b) above, the machine receives as input the test information of a plurality of users acquired in the past, the survey information of a plurality of users acquired in the past, and the parameter data set in the device in the past. A fitting support method, comprising: a learning model generated by learning and used for estimating a matching index of the parameter data and its distribution. to the computer,
(a) obtaining test information representing results of tests performed on a subject and survey information representing results of surveys related to the subject;
(b) inputting the obtained inspection information and the survey information in the fitting of the parameter data used to adapt the device to the subject, and estimating a conformance index representing the degree of conformity of the parameter data and its distribution; do, step and
(c) if there is missing information in the inspection information used for fitting, or in the investigation information, or both, the amount of information representing the correlation between the missing information and the parameter data is calculating for the information that is missing, and selecting the missing information based on the calculated amount of information;
program to run.

Images