Eating habits estimation device

Abstract

The user terminal (4) comprises: an analysis unit (42) (a chewing interval determination module (42a) and a filtering / calculation module (42b)) for obtaining chewing information, which is time series information representing the jaw movement of a bite amount when the user eats; and an analysis unit (42) (an analysis module (42c)) for inputting the chewing information into an analysis model (M1) and obtaining the attributes of food that is estimated to be chewed by the user through the jaw movement of the bite amount output from the analysis model (M1). The analysis model (M1) is a learned model obtained by machine learning using training data, and the training data includes first information, which is time series information representing the jaw movement of a past bite amount obtained for the user, and second information representing the attributes of food chewed by the user through the jaw movement of the past bite amount.

Description

Technical Field

[0001] The present invention relates to a dietary lifestyle estimation device. Background Art

[0002] Conventionally, there is known a device that uses information obtained from an intraoral sensor provided in the oral cavity of an animal including a human to analyze the behavior of the animal (see, for example, Patent Document 1).

[0003] Prior Art Literature

[0004] Patent Literature

[0005] Patent Document 1: Japanese Patent Application Publication No. 2012-191994 Summary of the invention

[0006] (Problems to be solved by the invention)

[0007] Here, if information about the food ingested (chewed) by the user can be obtained from the jaw movement of the user when eating as an analysis result, it is possible to provide the user with, for example, dietary advice based on the analysis result. On the other hand, in the method described in the above-mentioned Patent Document 1, various information for comparison with the information obtained from the intraoral sensor (information related to oral movements, information related to vibrations generated in the oral cavity, information related to vibrations generated in the biological body, etc.) is required in advance. It is difficult and time-consuming to obtain and accumulate such various information in advance.

[0008] Therefore, an object of the present invention is to provide a dietary lifestyle estimating device that can easily and accurately acquire information on food ingested by a user.

[0009] (Technical solutions to solve problems)

[0010] The dietary lifestyle estimation device involved in one aspect of the present invention includes an analysis unit, which obtains time series information of jaw movement representing a bite amount of a user when eating, namely chewing information, inputs the chewing information into an analysis model generated by machine learning, and obtains the attributes of food output from the analysis model, which is estimated to be chewed by the user through the jaw movement of one bite. The analysis model is a learned model obtained by machine learning using training data, and the training data includes first information, namely time series information of jaw movement representing past bite amounts obtained for the user, and second information representing the attributes of food chewed by the user through the jaw movement of one bite amount in the past.

[0011] In the above-mentioned dietary lifestyle estimation device, the attributes of the food estimated to be chewed by the user are obtained by using the analytical model generated by machine learning. In addition, the analytical model uses training data obtained from the past meals of the user himself who is the analysis object for machine learning. Therefore, according to the above-mentioned dietary lifestyle estimation device, it is possible to obtain the analysis result (the attributes of the food estimated to be chewed by the user) simply and with high accuracy using the analytical model that has learned the characteristics of the jaw movement of the user when eating.

[0012] The dietary lifestyle estimation device involved in another aspect of the present invention includes an analysis unit, which obtains time series information of jaw movement representing a bite amount of a user when eating, namely chewing information, inputs the chewing information into an analysis model generated by machine learning, and obtains the attributes of food output from the analysis model, which is estimated to be chewed by the user through the jaw movement of one bite. The analysis model is a learned model obtained by machine learning using training data, and the training data includes first information, namely time series information of jaw movement representing past bite amounts obtained for an unspecified user, and second information representing the attributes of food chewed by the unspecified user through the jaw movement of one bite amount in the past.

[0013] In the above-mentioned dietary lifestyle estimation device, the properties of food estimated to be chewed by the user are obtained by using an analytical model generated by machine learning. In addition, the analytical model uses training data obtained from past meals of unspecified users for machine learning. According to the above-mentioned dietary lifestyle estimation device, it is possible to use the analytical model to obtain analytical results (properties of food estimated to be chewed by the user) simply and with high accuracy. In addition, the analytical model can be made universal among multiple users, which has the advantage of not having to create and manage analytical models for each user. In addition, it also has the advantage of being easy to collect training data for the analytical model.

[0014] The analysis unit may also input profile information indicating the attributes of the user to the analysis model, and the analysis model is a learned model obtained by machine learning using training data that also includes profile information of an unspecified user. The jaw movement of the user when eating can be considered to be different depending on the attributes of the user (e.g., gender, age, health status, etc.). Therefore, according to the above configuration, by including the attributes of the user in the feature amount, the attributes of the food estimated to be chewed by the user can be obtained with higher accuracy.

[0015] The analytical model may be a learned model obtained by machine learning using training data, the training data including data corresponding to first chewing information, which is time series information indicating jaw movement corresponding to the first biting action in the first information, data corresponding to second chewing information, which is time series information indicating jaw movement corresponding to the second biting action in the first information, and second information, and the analytical unit extracts the first chewing information and the second chewing information from the chewing information and inputs the extracted information into the analytical model to obtain the properties of the food estimated to be chewed by the user through the jaw movement of one bite. In particular, features corresponding to the properties of the chewed food are likely to appear in the jaw movement corresponding to the first biting action and the second biting action. Therefore, according to the above structure, by using the analytical model that has been machine-learned using the jaw movement corresponding to the first biting action and the second biting action as feature quantities, it is possible to obtain the properties of the food estimated to be chewed by the user with higher accuracy.

[0016] The dietary lifestyle estimation device may further include an acquisition unit that acquires jaw movement information detected by a sensor provided on a denture worn on the user's lower jaw, and the analysis unit acquires chewing information based on the jaw movement information. Based on the result, the jaw movement information that is the basis of the chewing information can be acquired by the sensor provided on the denture worn on the user's lower jaw on a daily basis. In addition, by utilizing the sensor provided on the denture worn by the user in daily life including when eating, the jaw movement information can be appropriately acquired without placing an excessive burden on the user.

[0017] The jaw movement information may also include information indicating the time variation of at least one of the acceleration in the three-axis direction and the angular velocity in the three-axis direction detected by the sensor. In addition, the jaw movement information may also include information indicating the time variation of at least one of the acceleration in the three-axis direction and the angular velocity in the three-axis direction detected by the sensor provided on the denture worn on the upper jaw of the user. For example, when the user is riding in a vehicle such as a car or a train, the vibration component from the vehicle is detected by the sensor and may be mistakenly detected as the jaw movement of the user when eating. In addition, when the user eats while riding in a vehicle, the vibration component from the vehicle may be mixed into the jaw movement of the user when eating as noise. On the other hand, according to the above structure, the analysis unit can use not only the detection results of the sensor provided on the denture worn on the lower jaw of the user, but also the detection results of the sensor provided on the denture worn on the upper jaw of the user, and only obtain the movement component indicating the jaw movement of the user as the chewing information. For example, the analysis unit can eliminate the vibration component from the vehicle (i.e., the component contained in both the upper and lower jaw sensors) by obtaining the relative value of the detection result of the lower jaw sensor relative to the detection result of the upper jaw sensor. As a result, the above-mentioned problem can be eliminated.

[0018] In the dietary lifestyle estimation device, the attribute of the food may include at least one of the size, hardness, and type of the food.

[0019] The dietary lifestyle estimation device may also be configured such that the base device stores the denture removed from the user's lower jaw, and the base device acquires jaw movement information from the sensor by communicating with the sensor. According to the above structure, the jaw movement information can be automatically sent from the base device to the device that analyzes the jaw movement information simply by the user placing the denture on the base device. Thus, the user's convenience can be improved in acquiring (extracting) the jaw movement information from the sensor provided on the denture.

[0020] The base device may also be configured to further include a charging unit for charging the sensor. According to the above configuration, charging by the charging unit can be performed together with communication by the communication unit. Thus, the power of the sensor consumed in the communication operation by the communication unit can be appropriately supplemented.

[0021] The base device may also be configured to include a cleaning unit for cleaning the denture. According to the above structure, the user can simultaneously send the jaw movement information to the device for analyzing the jaw movement information and clean the denture simply by placing the denture on the base device. This can improve user convenience.

[0022] (Effects of the Invention)

[0023] According to the present invention, it is possible to provide a dietary lifestyle estimating device capable of acquiring information on food ingested by a user simply and with high accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 This is a block diagram showing the configuration of a dietary lifestyle estimation device according to one embodiment.

[0025] Figure 2 This is a block diagram showing the schematic structure of a denture and a base device.

[0026] Figure 3 It is a diagram schematically showing the outline of the analytical model.

[0027] Figure 4 is a diagram schematically showing an outline of a proposed model.

[0028] Figure 5 It is a diagram showing an example of processing by the analyzing unit and the generating unit.

[0029] Figure 6 This is a diagram showing an example of the chewing interval determined by the chewing interval determination module.

[0030] Figure 7 This is a diagram showing an example of chewing information for one bite.

[0031] Figure 8 This is a diagram showing an example of the sections corresponding to the first and second biting actions in the chewing information of a bite amount.

[0032] Fig. 9 (A) is a graph showing the simulation results when chewing almonds, (B) is a graph showing the simulation results when chewing rice, and (C) is a graph showing the simulation results when chewing strawberries.

[0033] Fig.10 This is a diagram showing an example of the processing flow of the dietary lifestyle estimation device.

[0034] Fig.11 : is a block diagram showing the structure of a dietary lifestyle estimation device according to a modification.

[0035] Fig.12 It is a diagram schematically showing an outline of an analytical model according to a modification example.

[0036] Fig.13 : is a block diagram showing the structure of a dietary lifestyle estimation device according to a modification. DETAILED DESCRIPTION

[0037] Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings. In each of the drawings, the same or corresponding parts are denoted by the same reference numerals, and duplicate descriptions are omitted.

[0038] [Overall Structure of the Dietary Life Estimation Device]

[0039] like Figure 1 As shown, a dietary lifestyle estimation device 1 (dietary lifestyle estimation system) according to one embodiment includes: a denture 2, a base device 3 (base 3), a user terminal 4, and a server 5. The denture 2 is a base denture that is worn on the lower jaw of a denture wearer (hereinafter referred to as a "user"). The denture 2 has a function of acquiring measurement data (jaw movement information) related to the jaw movement of the user when eating. The base device 3 is a device that stores the denture 2 removed from the lower jaw of the user. The user terminal 4 is a terminal used by the user. The user terminal 4 is, for example, a portable terminal such as a smartphone or a tablet owned by the user. Alternatively, the user terminal 4 may also be a personal computer or the like set up at the user's home, etc. The user terminal 4 mainly has a function of analyzing the measurement data acquired by the denture 2, and a function of generating food-related suggestion information for the user based on the analysis results. The server 5 is a device that generates an analysis model used when the user terminal 4 analyzes the measurement data and a suggestion model used when the user terminal 4 generates suggestion information, and provides the generated analysis model and suggestion model to the user terminal 4. In this embodiment, the artificial tooth 2 and the base device 3 constitute an actual measurement device MS for actually measuring the jaw movement of the user. In addition, the user terminal 4 and the server 5 constitute an operation device AS for performing analysis (operation) based on the measurement data.

[0040] [Structure of denture]

[0041] like Figure 1 As shown in FIG. 1 , the artificial tooth 2 has a base portion 2a mounted on the lower jaw of the user, and a tooth portion 2b (teeth) detachably provided with respect to the base portion. Figure 2 As shown, the denture 2 has a built-in sensor 20. The sensor 20 is composed of, for example, a circuit such as a processor (such as a CPU, etc.) and a memory (such as a ROM and a RAM, etc.) mounted on a substrate not shown in the figure. The sensor 20 includes: a sensor unit 21, a computing unit 22, a storage unit 23, a communication unit 24, a battery 25, and a wireless charging antenna 26. As an example of a configuration, the computing unit 22 is composed of a circuit such as a processor (such as a CPU, etc.) and a memory (such as a ROM and a RAM, etc.) mounted on a substrate not shown in the figure, and the storage unit 23 is composed of electronic storage elements such as a memory chip.

[0042] The sensor unit 21 includes various sensors for detecting the movement of the upper and lower jaws or the movement of the lower jaw of the user. In the detection of the movement of the upper and lower jaws or the lower jaw, a sensor that detects at least one of the acceleration and angular velocity of the movement of the upper and lower jaws or the lower jaw can be used, that is, at least one of an acceleration sensor and an angular velocity sensor can be used. As the above-mentioned acceleration sensor, a sensor that can measure the acceleration in the three-axis direction is preferred. In addition, in the movement of the lower jaw, there are many cases where rotation elements other than simple one-axis rotation are mixed, so for the above-mentioned angular velocity sensor, a sensor that can measure the angular velocity in the three-axis direction is also preferred. In this embodiment, the sensor unit 21 includes an acceleration sensor that detects the acceleration in each of the three-axis directions (X-axis direction, Y-axis direction, and Z-axis direction). In this case, the acceleration corresponding to the movement of the lower jaw (jaw movement) of the user is continuously detected by the sensor unit 21. However, the structure of the sensor unit 21 is not limited to the above. For example, the sensor unit 21 may also include an angular velocity sensor (gyro sensor) that detects the angular velocity in each of the three-axis directions instead of the above-mentioned acceleration sensor, and may also include both the above-mentioned acceleration sensor and the above-mentioned angular velocity sensor.

[0043] The operation unit 22 obtains the measurement data continuously obtained by the sensor unit 21 (in the present embodiment, the three-axis acceleration data indicating the acceleration in each of the three-axis directions), and stores the measurement data in the storage unit 23. In the present embodiment, the measurement data is time series information indicating the time change of the acceleration (the angular velocity when the sensor unit 21 includes an angular velocity sensor) corresponding to the jaw movement of the user. The storage unit 23 is constituted by, for example, the above-mentioned memory. However, the measurement data is not limited to the above-mentioned example, and for example, it may be time series information indicating the time change of the speed of the jaw movement. In addition, the operation unit 22 may also have a timing function. For example, in the case where the operation unit 22 is constituted by a processor or the like, it may be constituted as a built-in timer / counter in a manner having multiple processors. The timer / counter can be used for the time allocation of the time series information.

[0044] The calculation unit 22 adds the acquisition time of the measurement data to the measurement data at each time point acquired by the sensor unit 21, for example, by using the above-mentioned timer / counter. Then, the calculation unit 22 stores the measurement data with the acquisition time added in the storage unit 23. Thus, the storage unit 23 stores time series information in which the measurement data at each time point are arranged in the order of the acquisition time.

[0045] Furthermore, when the denture 2 is removed from the user's lower jaw and placed in the storage space 34 provided in the base device 3, the operation unit 22 performs processing corresponding to the control signal transmitted from the base device 3. For example, the operation unit 22 reads the measurement data stored in the storage unit 23 and transmits it to the base device 3 via the communication unit 24, or deletes the measurement data stored in the storage unit 23.

[0046] In addition, the operation unit 22 may control the operation mode of the sensor unit 21 as follows. For example, the operation unit 22 stores in advance the acceleration pattern corresponding to the user's eating action. Such a pattern may also be stored in the storage unit 23, for example. Moreover, the operation unit 22 may operate the sensor unit 21 in a power saving mode with a relatively low sampling rate (i.e., a relatively long sampling period) as an initial mode. In addition, the operation unit 22 may always monitor the measurement data continuously acquired by the sensor unit 21, and, when a pattern of acceleration corresponding to the user's eating action (e.g., acceleration above a predetermined threshold value, etc.) is detected as a trigger, the sensor unit 21 operates in a data acquisition mode with a sampling rate higher than the power saving mode (i.e., a short sampling period). Afterwards, the operation unit 22 may, for example, switch the operation mode of the sensor unit 21 to the power saving mode again when a predetermined set time (e.g., 30 seconds, etc.) has passed since the time point when the acceleration pattern corresponding to the user's eating action is no longer detected.

[0047] In this way, by setting the action mode of the sensor unit 21 to the power saving mode during the period other than when the user is eating, the power consumption of the sensor unit 21 can be reduced. As a result, even if the user continues to wear the denture 2 for a long time, the measurement data can be continuously acquired. In addition, by setting the action mode of the sensor unit 21 to the data acquisition mode when the user is eating, it is possible to acquire highly accurate measurement data representing the user's jaw movement when the user is eating.

[0048] Alternatively, the computing unit 22 may store the user's scheduled meal time periods (e.g., time periods corresponding to breakfast, lunch, snacks, and dinner, respectively), and operate the sensor unit 21 in the power saving mode in time periods other than the scheduled meal time periods, and operate the sensor unit 21 in the data acquisition mode in the scheduled meal time periods. By controlling such an operation mode, the same effect as the above effect can be obtained. In addition, in the above control related to the scheduled meal time, the above-mentioned timer / counter, etc. can be used.

[0049] The communication unit 24 has a function of communicating with the base device 3 through wireless communication when the denture 2 is placed in the storage space 34 of the base device 3. The communication unit 24 is composed of, for example, an antenna for sending and receiving information. Wireless communication is, for example, radio wave communication, optical communication, and sound wave communication. Radio wave communication is, for example, communication using Bluetooth (registered trademark), Wi-Fi (registered trademark), etc. Optical communication is, for example, communication using visible light or infrared rays. Sound wave communication is, for example, communication using ultrasonic waves, etc. If the communication unit 24 receives a control signal from the base device 3, it sends the control signal to the operation unit 22. In addition, if the communication unit 24 receives measurement data from the operation unit 22, it sends the measurement data to the base device 3.

[0050] The battery 25 is a storage battery that supplies power to the sensor unit 21, the computing unit 22, the storage unit 23, and the like. The wireless charging antenna 26 is a component for charging the battery 25. When the denture 2 is placed in the storage space 34 of the base device 3, power is supplied to the battery 25 from the base device 3 via the wireless charging antenna 26. That is, the battery 25 is charged by wireless charging. As described above, in the denture 2 of the present embodiment, since the battery 25 can be charged in a contactless manner, there is no need to expose the circuit (charging terminal) for charging the battery 25 to the outside of the denture 2.

[0051] The communication operation with the base device 3 based on the communication unit 24 can also be performed together with the charging operation of the battery 25. As described above, in the present embodiment, the above communication operation and the above charging operation are performed in a state where the denture 2 is placed in the storage space 34 of the base device 3. Here, the communication operation based on the communication unit 24 consumes corresponding power regardless of which of the above wireless communication methods is used, but when the charging operation is performed simultaneously with the communication operation, the power required for the communication operation can be appropriately supplemented.

[0052] At least a part of each part of the sensor 20 (especially the sensor part 21 and the battery 25 which are frequently replaced) may be embedded in a specific tooth part 2b. In this case, the electrical connection between each part of the sensor 20, that is, the electrical connection between the part embedded in the tooth part 2b (for example, the battery 25 embedded in the individual tooth part 2b1 in the tooth part 2b) and the other part (the part embedded in the base part 2a or the part embedded in other tooth parts 2b), may be ensured at the boundary between the tooth part 2b and the base part 2a. According to the above structure, the replacement operation of each part of the sensor 20 becomes easy. For example, the sensor part 21 can be easily replaced by simply removing the tooth part 2b (for example, the individual tooth part 2b2) in which the sensor part 21 is embedded from the base part 2a and attaching a new tooth part 2b (including the tooth part (individual tooth part 2b2) of the new sensor part 21) to the base part 2a.

[0053] [Structure of the base device]

[0054] like Figure 2 As shown, the base device 3 includes: a wireless charging antenna 31, a communication unit 32, and a cleaning unit 33 (cleaner 33). The base device 3 is installed in the user's home, etc. In view of the storage function of the base device 3, the base device 3 also includes a storage space 34 for storing the denture 2.

[0055] The wireless charging antenna 31 wirelessly charges the battery 25 of the denture 2 when the denture 2 is placed in the storage space 34 of the base device 3. Specifically, the wireless charging antenna 31 supplies power to the battery 25 via the wireless charging antenna 26 described above. The wireless charging antenna 31 constitutes a charging unit (charger). Specifically, the wireless charging antenna 31 constitutes a charging unit 31A on the side of the base device 3, and the wireless charging antenna 26 constitutes a charging unit 26A on the side of the sensor 20. The charging unit 31A and the charging unit 26A constitute a charging unit CH.

[0056] The communication unit 32 has a function of communicating with the sensor 20 of the denture 2 and the user terminal 4. Specifically, the communication unit 32 communicates with the communication unit 24 of the denture 2 by wireless communication when the denture 2 is placed in the storage space 34 of the base device 3. In addition, the communication unit 32 communicates with the user terminal 4 by wired communication or wireless communication. For example, when the communication unit 32 receives a control signal indicating a request for measurement data from the user terminal 4, the communication unit 32 sends the control signal to the communication unit 24 of the denture 2 placed on the base device 3. Then, when the communication unit 32 receives measurement data from the communication unit 24, the communication unit 32 sends the measurement data to the user terminal 4. Thus, the measurement data acquired by the sensor 20 of the denture 2 is sent to the user terminal 4 via the base device 3. Then, in the user terminal 4, the measurement data is analyzed, and proposal information corresponding to the analysis result is generated. The communication unit 32 is a communication unit on the side of the base device 3, and the communication unit 24 is a communication unit on the side of the sensor 20. The communication unit 32 and the communication unit 24 constitute the communication unit TR.

[0057] The cleaning unit 33 has a function of cleaning the denture 2 when the denture 2 is placed in the storage space 34 of the base device 3. The cleaning unit 33 performs, for example, a process of introducing a liquid containing a dedicated agent into the storage space 34, a process of cleaning the surface of the denture 2 by immersing the denture 2 in the liquid, and a process of extracting the liquid from the storage space 34 after cleaning and drying the denture 2. That is, the storage space 34 also serves as a cleaning space capable of cleaning the denture 2.

[0058] As described above, the base device 3 has the functions of storing the denture 2, charging the sensor 20, communicating with the sensor 20 (acquiring measurement data from the sensor 20, transmitting control signals to the operation unit 22, etc.), and cleaning the denture 2. Thus, the user can store the denture 2, charge the sensor 20, acquire measurement data from the sensor 20 (and transmit to the user terminal 4), and clean the denture 2, simply by removing the denture 2 from the user's lower jaw before going to bed, for example, and placing the denture 2 in the storage space 34 of the base device 3. According to such a base device 3, measurement data related to the jaw movement of the user in daily life can be acquired without placing an excessive burden on the user.

[0059] [Structure of user terminal]

[0060] like Figure 1As shown, the user terminal 4 includes: an acquisition unit 41, an analysis unit 42, an analysis model storage unit 43, a generation unit 44, a proposed model storage unit 45, an output unit 46, and a display unit 47. The user terminal 4 is a computer device including a processor (such as a CPU, etc.) and a memory (such as a ROM and a RAM, etc.). In addition, the user terminal 4 may also include a timer / counter like the calculation unit 22 of the sensor 20. The acquisition unit 41, the analysis unit 42, the generation unit 44, the output unit 46, and the display unit 47 may be respectively constituted by circuits, or part or all of their circuits may be constituted by a processor. The acquisition unit 41, the analysis unit 42, the generation unit 44, the output unit 46, and the display unit 47 may also respectively constitute part or all of the processor, and each circuit may also include a memory.

[0061] The acquisition unit 41 acquires measurement data related to the jaw movement of the user when eating. As described above, the acquisition unit 41 acquires the measurement data acquired and stored in the sensor 20 of the denture 2 via the base device 3. However, the method by which the acquisition unit 41 acquires the measurement data is not limited to the above. For example, the acquisition unit 41 may also acquire the measurement data by directly communicating with the sensor 20 (communication unit 24) without going through the base device 3.

[0062] In this embodiment, the measurement data is time series information (three-axis acceleration data) indicating time changes in acceleration in each of the three-axis directions detected by the sensor 20 provided in the denture 2 worn on the user's lower jaw. The three-axis acceleration data is information indicating the X-axis component, the Y-axis component, and the Z-axis component of the acceleration at each moment continuously acquired at a given sampling period in the sensor unit 21.

[0063] For example, when the user places the denture 2 in the storage space 34 of the base device 3 while sleeping, the acquisition unit 41 can acquire the user's daily measurement data every night (i.e., the period from when the user gets up and puts on the denture 2 to when the denture 2 is removed before going to bed). Here, the acquisition unit 41 can acquire all the measurement data stored in the storage unit 23 of the sensor 20 (i.e., the daily measurement data), or it can only acquire the measurement data acquired by the sensor unit 21 in the data acquisition mode (i.e., the user's measurement data when eating). For example, the acquisition unit 41 can also send a control signal requesting only the measurement data acquired in the data acquisition mode to the sensor 20 (the calculation unit 22) via the base device 3, and acquire only the measurement data acquired in the data acquisition mode from the sensor 20. According to the latter, it is possible to efficiently acquire only the data required for the analysis of the analysis unit 42 described later (i.e., the user's measurement data when eating). By omitting the communication of unnecessary measurement data in this way, the data communication volume can be reduced and the measurement data required for analysis can be acquired in a short time.

[0064] The analysis unit 42 performs analysis based on the measurement data obtained by the acquisition unit 41 to obtain the analysis result of the user's eating status when eating. In this embodiment, the analysis result includes attribute information indicating the attributes of the food that is estimated to be chewed (ingested) by the user when eating. The so-called food attribute includes at least one of the size, hardness and type of the food. The so-called type of food refers to information that determines foods such as almonds, rice (rice), strawberries, etc. In this embodiment, the analysis result includes: all attribute information including the size, hardness and type of the food estimated to be chewed by the user, and the estimated value of the user's masticatory muscle activity when eating. In addition, in addition to the above, the analysis result may also include various information obtained by analyzing the measurement data using statistical methods, etc. (for example, information such as the number of chewings per bite, the bite force per bite, and the posture of each bite (the posture of the head when eating)). In addition, the attribute information of the estimated food may also be information that at least includes hardness. Although the details are described later, the hardness property of the chewing object can be appropriately estimated by detecting the acceleration and / or angular velocity of the movement of the jaw during the first biting action and the second biting action of the chewing object. By being able to estimate the hardness of the chewing object, it becomes easy to make suggestions for improving the diet and lifestyle, including recommending the intake of harder foods.

[0065] The analysis unit 42 obtains part or all of the above analysis results from the measurement data by using the analysis model (analysis model M1) stored in the analysis model storage unit 43. In the present embodiment, the analysis unit 42 extracts (obtains) the chewing information (time series information of the jaw movement representing the user's mouthful) from the measurement data. Figure 7 as well as Figure 8 The analysis unit 42 then inputs the chewing information of a mouthful into the analysis model, and obtains information output from the analysis model as part of the above analysis result. In this embodiment, the information output from the analysis model includes: attribute information indicating the attributes of the food that is estimated to be chewed by the user through the jaw movement of a mouthful (in this embodiment, the size, hardness and type of the food), and the estimated value of the activity of the chewing muscles of a mouthful. The details of such an analysis model will be described later.

[0066] Here, the so-called jaw movement of a mouthful refers to the jaw movement from the time when the user opens his mouth to put a mouthful of food into the oral cavity and puts the food into the mouth, to the time when the user opens his mouth again to put the next mouthful of food into the oral cavity. Alternatively, since the user usually opens his mouth to put a mouthful of food into the oral cavity, chews the food after putting it into the mouth, and then opens his mouth again to put the next mouthful of food into the oral cavity after swallowing, the jaw movement of a mouthful can also be set as the jaw movement from the time when the user opens his mouth to put a mouthful of food into the oral cavity to the time when the chewing ends. That is, the so-called jaw movement of a mouthful refers to the jaw movement during the period when the user chews the mouthful of food without putting new food into the mouth after putting a mouthful of food into the oral cavity. The jaw movement of a mouthful can also be considered as the jaw movement during the period from the time when the user opens his mouth to chew a mouthful of food to the time when the user opens his mouth to chew the next mouthful of food during a meal, which at least includes the period from the start to the end of chewing the mouthful of food put into the oral cavity. A single meal usually involves repeated chewing of a mouthful of food, and thus can be defined as described above. Alternatively, chewing can be determined to be complete when a certain period of time has passed without jaw movement after the chewing is complete, so that information on the jaw movement of the last mouthful can be processed. In addition, the jaw movement of a mouthful can also be considered to be the jaw movement during chewing from opening the mouth to chewing a mouthful of food until swallowing the mouthful of food placed in the mouth without placing new food in the mouth.

[0067] The generating unit 44 generates the suggestion information related to the user's eating based on the analysis result obtained by the analyzing unit 42. The suggestion information is, for example, a suggestion on the content of the meal (e.g., a reminder to chew less, not eat rice, etc.), an evaluation of the tendency of recent meals (evaluation compared with the standard diet), a suggestion for making the diet better (e.g., it is better to increase the number of chews, it is better to fix the meal time, it is better to extend the meal time, it is better to increase the frequency of consuming harder foods because you have not eaten much hard food recently, etc.). The suggestion information can be, for example, a template selected from a plurality of templates intended in advance (e.g., data containing text corresponding to the suggestion content), or it can be a combination of several templates (i.e., a composite suggestion content). The suggestion on the content of the meal or the suggestion for making the diet better can be said to be a suggestion for improving the eating behavior (direct suggestion for improving the eating behavior) that directly suggests a better eating behavior to the user. The evaluation of the tendency of recent meals also has the effect of promoting motivation to improve when the evaluation is poor, so it can be said to be a suggestion for improving the eating behavior (indirect suggestion for improving the eating behavior) that indirectly encourages the user to have a better eating behavior.

[0068] The generation unit 44 generates (acquires) part or all of the above-mentioned proposal information based on part or all of the analysis results obtained by the analysis unit 42 by using the proposal model (proposal model M2) stored in the proposal model storage unit 45. As an example, the generation unit 44 inputs the analysis results corresponding to a given period into the proposal model, and obtains the information output from the proposal model as part of the above-mentioned proposal information. Here, the so-called given period is, for example, a meal period for one meal, a meal period for one day (that is, a period including breakfast, lunch, snacks, and dinner), etc. The details of such a proposal model will be described later.

[0069] The output unit 46 outputs the proposal information generated by the generating unit 44. In the present embodiment, the output unit 46 outputs the proposal information to the display unit 47. Specifically, the output unit 46 generates display data for displaying the content of the proposal information on the screen, and outputs the display data to the display unit 47.

[0070] The display unit 47 displays the suggestion information. The display unit 47 is constituted by, for example, a display device provided in the user terminal 4. The display unit 47 displays the display data obtained from the output unit 46, thereby presenting the content of the suggestion information to the user. Thus, the user can confirm the suggestion content corresponding to the analysis result of the measurement data obtained by the sensor 20 of the denture 2.

[0071] The following describes a mechanical structure example of the acquisition unit 41, the analysis unit 42, the analytical model training data storage unit 52, the analytical model storage unit 43, the generation unit 44, and the output unit 46. The acquisition unit 41 is composed of an interface for acquiring measurement data. When the measurement data is processed, the acquisition unit 41 is composed of an interface and a circuit. The circuit may also be composed of a part or all of a processor. The analysis unit 42 is composed of a program for performing analysis and a circuit for executing the program. The circuit may also be composed of a part or all of a processor. The program for performing analysis is composed of an electrical signal stored in a memory. The analytical model training data storage unit 52 is composed of a memory. The analytical model storage unit 43 is composed of a memory. The generation unit 44 is composed of a program for performing generation and a circuit for executing the program. The circuit may also be composed of a part or all of a processor. The program for performing generation is composed of an electrical signal stored in a memory. The output unit 46 is composed of a program for performing output and a circuit for executing the program. The circuit may also be composed of a part or all of a processor. The program for performing output is composed of an electrical signal stored in a memory.

[0072] [Server Structure]

[0073] like Figure 1As shown, the server 5 has a training data generation unit 51 for an analytical model, a training data storage unit 52 for an analytical model, and an analytical model learning unit 53 as functional elements for generating the above-mentioned analytical model. In addition, the server 5 has a training data generation unit 54 for a proposed model, a training data storage unit 55 for a proposed model, and a proposed model learning unit 56 as functional elements for generating the above-mentioned proposed model. The server 5 is a computer device having a processor (such as a CPU, etc.) and a memory (such as a ROM and a RAM, etc.). The server 5 can be composed of one or more devices set up on the cloud, for example.

[0074] The analytical model training data generation unit 51 generates training data for learning an analytical model. The analytical model training data storage unit 52 stores the training data generated by the analytical model training data generation unit 51. The analytical model learning unit 53 generates an analytical model as a learned model by executing machine learning using the training data stored in the analytical model training data storage unit 52.

[0075] Figure 3 is a diagram schematically showing an overview of the analytical model. Figure 3 As shown, first, the analytical model training data generation unit 51 generates multiple sets of training data, and the training data includes time series information representing the jaw movement of the past bite obtained for the user, that is, chewing information (first information) and information on the actual meal content of the user corresponding to the jaw movement of the past bite. The information on the actual meal content of the user is information on the properties of the food actually chewed by the user through the jaw movement of the past bite (second information). In this embodiment, as information on the actual meal content, the amount of masticatory muscle activity corresponding to the jaw movement of the past bite is also included. The masticatory muscle activity is measured during the user's actual meal. In addition, in addition to the above, the information on the actual meal content may also include the results of measuring the bite force during the eating action, the head posture, etc.

[0076] Such training data can be obtained, for example, by having the user eat various foods and establishing a correspondence between the chewing information of each bite extracted from the measurement data at that time and the user's actual meal content. However, the method of obtaining training data is not limited to the above. Training data can also be obtained, for example, by having the user wear smart glasses that can record images of the user's mouth, and using the smart glasses to record the user's actual meal pattern. Specifically, the chewing information of each bite of the meal is obtained from the measurement data obtained by the sensor 20 of the denture 2, and the user's actual meal content is grasped from the image obtained by the smart glasses. Then, by establishing a correspondence between the user's actual meal content grasped based on the image obtained by the smart glasses and the chewing information for each bite, the above-mentioned training data can be obtained. The training data generated in this way is accumulated in the training data storage unit 52 for the analytical model.

[0077] Next, the analytical model learning unit 53 generates the analytical model M1 as a learned model by executing machine learning using the training data accumulated in the analytical model training data storage unit 52. The method of machine learning executed by the analytical model learning unit 53 is not limited to a specific method, and various methods such as SVM, neural network, and deep learning can be used. For example, when the analytical model M1 is composed of a neural network, a learned model in which the parameters of the intermediate layer of the neural network are adjusted by the training data is obtained as the analytical model M1.

[0078] The analytical model M1 obtained by such machine learning is a learned model configured to input the chewing information of a user's bite size and output an estimated result of the user's meal content corresponding to the chewing information of the bite size. Here, the estimated result of the user's meal content includes the estimated result of the attributes of the food such as the size, hardness and type of the food in a bite size, and the estimated value of the masticatory muscle activity in a bite size. The analytical model M1 generated by the analytical model learning unit 53 is provided from the server 5 to the user terminal 4 and stored in the analytical model storage unit 43 of the user terminal 4.

[0079] The proposed model training data generation unit 54 generates training data for learning the proposed model. The proposed model training data storage unit 55 stores the training data generated by the proposed model training data generation unit 54. The proposed model learning unit 56 generates a proposed model as a learned model by performing machine learning using the training data stored in the proposed model training data storage unit 55.

[0080] Figure 4 is a diagram schematically showing an overview of the proposed model. Figure 4As shown, first, the training data generation unit 54 for the proposed model generates a plurality of sets of training data, the training data including the analysis results corresponding to the given period in the past obtained for the user, and the proposed information related to the food intake corresponding to the analysis results. Here, as an example, the given period is a meal period (for example, breakfast, etc.). In addition, the analysis results corresponding to the given period may include the attribute information (size, hardness and type) of the food chewed by the user during the given period, the activity of the masticatory muscles during the given period, the total number of chewing times during the given period, the bite force during the given period, the posture of the head during the food intake action during the given period, etc. The bite force may be actually measured, or may be analyzed based on the average value of a standard based on, for example, gender, age, physique, etc. That is, the analysis results corresponding to the given period input to the proposed model include not only the information obtained using the above-mentioned analysis model M1 (the attribute information of the food chewed by the user and the estimated value of the activity of the masticatory muscles, etc.), but also the information obtained by statistical methods, etc. without using the analysis model M1. In addition, the posture of the head may also be measured based on the detection result of the sensor unit 21.

[0081] The proposed information related to eating included in the above-mentioned training data may also be the above-mentioned proposal for improving eating behavior. Alternatively, the proposed information may also be, for example, proposed information corresponding to the user's health status (e.g., cold, stomachache, headache, etc.) during the above-mentioned given period in the past. The training data generated by the proposed model training data generation unit 54 is accumulated in the proposed model training data storage unit 55. The user's health status can be considered to be reflected in the user's dining actions. Although it is believed that each user has individual differences, for example, when the user suffers from a cold, stomachache, etc., it is possible that the amount of food consumed is less than usual, or the food consumed is biased, or the bite force is smaller than usual. In addition, when the user suffers from a headache, the posture of the head may also be worse than usual (e.g., tilted). In this way, it can be said that there is a certain correlation between the user's health status and the user's dining actions. Therefore, it can be said that there is also a certain correlation between the proposed information corresponding to the user's health status (i.e., the proposed information considered useful to the user based on the user's health status) and the user's dining actions. Therefore, according to the machine learning based on the above training data, it is possible to generate a proposal model that is configured to input the analysis result corresponding to a given period and output proposal information corresponding to the health status of the user estimated based on the analysis result. As an example of a proposal, in the case of suspected disease, it is possible to consider prompting the user of the cause and making appropriate menu proposals, advising the user to see a doctor, and other proposals related to healing support. Advice to see a doctor is a proposal (healthy proposal) for bringing the user to a better health status in association with food intake.

[0082] Next, the proposed model learning unit 56 generates a proposed model M2 as a learned model by executing machine learning using the training data accumulated in the proposed model training data storage unit 55. The method of machine learning executed by the proposed model learning unit 56 is not limited to a specific method, and various methods such as SVM, neural network, and deep learning can be used. For example, in the case where the proposed model M2 is composed of a neural network, a learned model in which the parameters of the intermediate layer of the neural network are adjusted by the training data is obtained as the proposed model M2.

[0083] The proposal model M2 obtained by such machine learning is a learned model configured to input the analysis result corresponding to the given period of the user and output the proposal information corresponding to the analysis result. The proposal model M2 generated by the proposal model learning unit 56 is provided from the server 5 to the user terminal 4 and stored in the proposal model storage unit 45 of the user terminal 4.

[0084] The following describes a mechanical configuration example of the analytical model training data generation unit 51, the analytical model learning unit 53, the proposed model training data generation unit 54, the proposed model training data storage unit 55, and the proposed model learning unit 56. The analytical model training data generation unit 51 is composed of a program for generating analytical model training data and a circuit for executing the program. The circuit may be composed of a part or the whole of a processor. The program for generating analytical model training data is composed of an electrical signal stored in a memory. The analytical model learning unit 53 is composed of a program for performing analytical model learning and a circuit for executing the program. The circuit may be composed of a part or the whole of a processor. The program for performing analytical model learning is composed of an electrical signal stored in a memory. The proposed model training data generation unit 54 is composed of a program for generating proposed model training data and a circuit for executing the program. The circuit may be composed of a part or the whole of a processor. The program for generating proposed model training data is composed of an electrical signal stored in a memory. The proposed model training data storage unit 55 is composed of a memory. The proposed model learning unit 56 is composed of a program for performing proposed model learning and a circuit for executing the program. The circuit may be formed by a part or all of the processor. The program for learning the proposed model is formed by electrical signals stored in the memory.

[0085] [Processing of the parsing unit and the generation unit]

[0086] Reference Figures 5 to 8 , an example of the processing of the analysis unit 42 and the generation unit 44 is described. Figure 5 4 is a diagram showing an example of a process flow from measurement data at the time of ingestion (measurement data acquired by the acquisition unit 41) to generation of suggestion information. Figure 5As shown, in this embodiment, the analysis unit 42 includes a chewing interval determination module 42a, a filtering / calculation module 42b, and an analysis module 42c. In addition, the generation unit 44 includes a suggestion module 44a. These modules are composed of programs for executing various processes. Figure 5 In the figure, the analysis module 42c is shown to have the analysis model M1. This means that the analysis module 42c uses the analysis model M1 in some form, such as when the analysis model M1 is stored in the memory of the analysis module 42c, when the analysis module 42c accesses the analysis model M1 in the analysis model storage unit 43, and when the analysis model M1 is taken from the analysis model storage unit 43 to the memory of the analysis module 42c.

[0087] [Mastication interval determination module]

[0088] First, the analysis unit 42 extracts the 3-axis acceleration data for each mouthful from the measurement data (here, 3-axis acceleration data) of a certain period (for example, a daily amount from when the user puts on the denture 2 after getting up to when the user takes off the denture 2 before going to bed. The period from the fitting of the denture 2 to the removal of the denture 2 in the user's daily life can also be referred to as the denture wearing period) through the chewing interval determination module 42a. Specifically, the chewing interval determination module 42a determines the chewing interval for each mouthful contained in the measurement data (that is, the interval in which the user puts a mouthful of food into the mouth and performs chewing movements). The chewing interval for each mouthful includes multiple biting movements (chewing movements) for crushing food in the oral cavity. Then, the chewing interval determination module 42a extracts the measurement data corresponding to each determined chewing interval as a mouthful of data.

[0089] Figure 6 : is a diagram showing an example of a chewing interval CS determined by the chewing interval determination module 42a. X, Y, and Z in the figure represent accelerations in the X-axis direction, the Y-axis direction, and the Z-axis direction, respectively. Here, it is known that the pattern of jaw movement of a certain user during chewing is not related to the type of food, etc. but is generally constant. By using this content, the chewing interval determination module 42a determines the chewing interval, for example, in the following manner.

[0090] First, the chewing interval determination module 42a obtains in advance the period of the user's jaw movement when chewing (hereinafter referred to as the "jaw movement cycle"). For example, the jaw movement cycle can be obtained (calculated) based on a prior preliminary measurement (for example, video observation of the user's chewing movement, etc.). In addition, for example, when there is an interval in the measurement data in which it is determined that the user is performing a chewing action, by extracting the periodic motion component observed in the interval, the period of the motion component can be obtained as the above-mentioned jaw movement cycle. In the jaw movement cycle, in addition to the actual movement of the active jaw, the maintenance movement of the inactive jaw is sometimes included. In the following, in the jaw movement cycle, the movement of the active jaw is represented as active jaw movement, and the maintenance movement of the inactive jaw is represented as passive jaw movement. Periodic jaw movement can be composed of active jaw movement, or it can be composed of a combination of active jaw movement and passive jaw movement. Therefore, the aforementioned jaw movement of one bite can be composed of active jaw movement, or it can be composed of a combination of active jaw movement and passive jaw movement.

[0091] Next, the chewing interval determination module 42a determines from the measured data the interval in which the periodic motion component is observed with a cycle that is consistent with or similar to the jaw movement cycle, as the chewing interval. For example, when it is detected that the periodic motion component continues for a predetermined threshold time (for example, a few seconds) or longer, the chewing interval determination module 42a determines the detected interval as the chewing interval. Then, the measured data corresponding to the chewing interval determined in this way is extracted as the data of the bite size. Through such processing by the chewing interval determination module 42a, as shown in FIG. Figure 6 As shown, the chewing interval CS is determined from the measurement data (here, the 3-axis acceleration data). In this way, the chewing interval CS of each bite is determined according to each meal period of the user contained in the measurement data (for example, breakfast, lunch, snack, dinner, etc.). That is, a plurality of chewing intervals CS are determined according to each meal period contained in the measurement data. Then, the measurement data contained in each chewing interval CS is extracted as the 3-axis acceleration data of each bite. The chewing interval determination module 42a can also extract the jaw movement used for chewing from the jaw movement of the user to determine the start period of chewing. In addition, the chewing interval determination module 42a can also determine that chewing is over when a certain period of time has passed without jaw movement after the end of chewing. In addition, the chewing interval determination module 42a can also divide the start and end of a meal, set the period of a meal as a meal interval, and determine the chewing interval CS in the meal interval.

[0092] [Filter / Calculation Module]

[0093] Next, the analysis unit 42 removes the noise components and the influence of gravity acceleration contained in the three-axis acceleration data of a bite through the filtering / calculation module 42b. In the present embodiment, the filtering / calculation module 42b removes the noise of the high-frequency components contained in the three-axis acceleration data of a bite through, for example, a low-pass filter, and performs offset correction on the three-axis acceleration data of a bite. The so-called offset correction refers to the process of removing the signal components caused by the gravity acceleration obtained by the acceleration sensor. Such a process of removing or reducing the noise components that may have an adverse effect on the analysis can also be called "adverse effect component countermeasure processing". The filtering / calculation module 42b can also be called an "adverse effect component countermeasure module" that performs adverse effect component countermeasure processing.

[0094] like Figure 7 As shown, the filtering / calculation module 42b obtains time series data D1 representing the norm of acceleration based on the three-axis acceleration data of a mouthful after noise removal and offset correction. Next, the filtering / calculation module 42b subtracts the gravitational acceleration (1g) from the norm of acceleration and integrates the subtraction result. Thus, time series data D2 is obtained. The time series data D2 becomes waveform data similar to the user's jaw movement (angular velocity of the lower jaw).

[0095] The time series data D1 and D2 are used as chewing information of one mouthful of the user to be input to the analysis model M1. That is, the analysis unit 42 extracts chewing information representing the jaw movement of one mouthful of the user, i.e., the time series data D1 and D2, from the measurement data acquired by the acquisition unit 41 through the chewing interval determination module 42a and the filtering / calculation module 42b. As described above, the chewing information (time series data D1 and D2) is extracted for each chewing interval CS determined by the chewing interval determination module 42a.

[0096] The time series data D1 and D2 can be used to obtain information about the size of the user's mouth opening (i.e., the size of the lower jaw opening relative to the upper jaw in order to bite the food) and the biting movement (biting action). For example, based on the characteristics of the waveforms of the time series data D1 and D2, it is possible to extract the interval from the state in which the user opens the mouth widely (i.e., the state in which the lower jaw is separated from the upper jaw), closes the mouth to bite the food, swallows it, and opens the mouth widely again to put the next bite of food into the mouth. This interval is the interval in which one biting action is performed (in this embodiment, the action of temporarily opening the mouth to bite the food and opening the mouth again for the next biting action).

[0097] Figure 8: This is a diagram showing the interval B1 corresponding to the first biting action and the interval B2 corresponding to the second biting action extracted from the time series data D1 and D2. Such biting motion of the first biting action (data in interval B1) and the biting motion of the second biting action (data in interval B2) are the parts that are particularly highly correlated with the type (type of food determined by hardness, cohesion, etc.) of food (chewed food), size, and masticatory muscle activity. In other words, the data in intervals B1 and B2 are the parts where the characteristics corresponding to the type, size, and masticatory muscle activity of food are likely to appear.

[0098] Fig. 9 An example of data obtained by the sensor 20 of the denture 2 when the robot is equipped with the denture 2 simulating the actual jaw movement of a person chewing is shown. Here, the sensor part 21 of the sensor 20 is provided with an angular velocity sensor for detecting the angular velocity of the rotational motion of the mandible around the axis connecting the reference point of the mandible in addition to the 3-axis acceleration sensor. As a specific example of such an angular velocity sensor, an angular velocity sensor for detecting the angular velocity of the rotational motion of the mandible around the axis connecting the left and right mandibular condylar points of the mandible can be cited. As the angular velocity sensor, a 3-axis angular velocity sensor is preferably used. When a robot is used, it is easy to set which part is used as the reference point. The time series data D3 is data obtained by integrating the norm of the angular velocity obtained from the data obtained by the 3-axis angular velocity sensor. That is, the time series data D3 is data corresponding to the angle of the mandible relative to the upper jaw at each time point. The time series data D4 is data corresponding to the above-mentioned time series data D2 (that is, data obtained by integrating the result obtained by subtracting the gravitational acceleration from the norm of the acceleration). Fig. 9 In the graph shown in FIG. 1 , the smaller the value of the time series data D3 (located at the bottom), the larger the mouth opening (the opening of the lower jaw relative to the upper jaw). Figures 5 to 8 In the example, the analysis unit 42 only obtains the chewing information of one bite (here, the time series data D1 and D2) from the three-axis acceleration data, but when the sensor unit 21 has an angular velocity sensor in the three-axis direction, the analysis unit 42 can also obtain data equivalent to the above-mentioned time series data D3 as the chewing information of one bite.

[0099] Fig. 9 (A) is a graph showing simulation results in the case of chewing almonds. Fig. 9 (B) is a graph showing the simulation result when chewing a mouthful of cooked rice (rice). Fig. 9 (C) is a graph showing the simulation result in the case of chewing strawberries. Here, the physical characteristics of each food are as follows.

[0100] Almonds: Smaller and harder.

[0101] Rice: Smaller and softer.

[0102] ·Strawberries: Larger and softer.

[0103] like Fig. 9 As shown, in the overall waveform of the time series data D3 and D4, differences between foods appear (i.e., differences corresponding to the differences in the above-mentioned physical characteristics). In particular, the differences between foods appear significantly in the portions corresponding to the first biting action (interval B1) and the second biting action (interval B2) in the time series data D3 and D4. For example, in the second biting action of the almonds, there appears a characteristic of further biting action after the movement of the jaw is temporarily stopped. In addition, for rice with high viscosity, the number of chewing times (number of biting actions) per unit time is greater than that of the other two. In addition, for strawberries, the size of the opening in the second and subsequent biting actions is relatively small. It can be inferred that this is because the food mass is moved while being eaten with the tongue.

[0104] Thus, it can be seen that the chewing information of the user's bite size has features corresponding to the physical features of the chewed food. In addition, although it is believed that there are individual differences in the jaw movement during chewing, as described above, by using a plurality of training data (correct answer data) obtained from the user's past meals to learn the analytical model M1, it is possible to obtain an analytical model M1 corresponding to the features of the jaw movement during chewing that are unique to the user.

[0105] The input data for the analytical model M1 may also include time series information (first chewing information) indicating the jaw movement corresponding to the first biting action and time series information (second chewing information) indicating the jaw movement corresponding to the second biting action. The first chewing information is the data included in the interval B1, and the second chewing information is the data included in the interval B2. That is, for the analytical model M1, the data included in the interval B1 may be explicitly input as the data corresponding to the first biting action, and the data included in the interval B2 may be explicitly input as the data corresponding to the second biting action. In this case, the analytical model training data generation unit 51 extracts the data corresponding to the first biting action (i.e., the information corresponding to the first chewing information) and the data corresponding to the second biting action (i.e., the information corresponding to the second chewing information) from the chewing information of the past bite amount obtained from the user, and includes the extracted data in the training data for the analytical model. In this way, it is preferable that at least the data corresponding to the first biting action and the data corresponding to the second biting action are included in the training data. In this case, the analytical model M1 is learned to output analytical results for the first chewing information and the second chewing information (e.g., attribute information of food estimated to be chewed by the user, estimated value of masticatory muscle activity of a mouthful, etc.). The data corresponding to the first biting action may be referred to as "first chewing information analytical education data," and the data corresponding to the second biting action may be referred to as "second chewing information analytical education data."

[0106] [Analysis module]

[0107] Next, the analysis unit 42 obtains the above-mentioned analysis result from the chewing information of a bite through the analysis module 42c. The analysis unit 42 inputs the chewing information of a bite into the analysis model M1 stored in the analysis model storage unit 43. As described above, the chewing information of a bite input to the analysis model M1 may include only the time series data D1 and D2 (hereinafter referred to as "overall interval data") in the chewing interval CS, or only the time series data D1 and D2 (hereinafter referred to as "specific interval data") included in the intervals B1 and B2 in the chewing interval CS, or both the overall interval data and the specific interval data. In addition, the chewing information of a bite input to the analysis model M1 may include only one of the time series data D1 and D2. In addition, as described above, when the chewing information of a bite includes the time series data corresponding to the above-mentioned time series data D3 (that is, data indicating the angle of the mandible), the time series data may be included in the chewing information of a bite input to the analysis model M1. In addition, the first and second bite actions are particularly focused on because they significantly show the characteristics of the chewed object, but it is also possible to focus on any other number of bite actions (nth time). That is, chewing information corresponding to the first to nth bite actions can also be used as input data for the analytical model M1. It goes without saying that the above n times is set to a number smaller than the total number of bite actions in the chewing interval CS.

[0108] Then, the analysis unit 42 obtains the attributes of the food (in this embodiment, the size, hardness and type of the food) output from the analysis model M1 and estimated to be chewed by the user through the jaw movement of one mouthful, as the above-mentioned attribute information through the analysis module 42c. That is, the analysis unit 42 (analysis module 42c) uses the analysis model M1 to perform the estimation of the attributes. The attribute information of the food output from the analysis model M1 can also be referred to as "estimated attribute information", and the attributes of the food estimated to be chewed can be referred to as "estimated attributes". In addition, as in the present embodiment, the analysis model M1 can also output the estimated value of the amount of masticatory muscle activity for one mouthful together with the attributes of the above-mentioned food. In addition, the analysis model M1 can also be configured to output information other than the above-mentioned information.

[0109] In addition, the analysis unit 42 may also obtain information obtained without using the analysis model M1 as part of the analysis result. The information obtained without using the analysis model M1 is, for example, information such as the bite force per bite, the posture of the user's head during the eating action, etc. As a method of obtaining information without using the analysis model M1, there are, for example, analyzing the measurement data using a statistical method, using the sensor unit 21 to obtain the information, etc.

[0110] For example, the posture of the user's head during food intake can be derived from the three-axis acceleration data. Specifically, the analysis unit 42 derives an estimation formula for estimating the posture of the user's head (face orientation) from the three-axis acceleration data in advance through the following process.

[0111] 1. The artificial tooth 2 is kept stationary so as to face the first direction, the second direction and the third direction which are orthogonal to each other.

[0112] 2. Obtain first acceleration data in three-axis directions measured by the acceleration sensor of the sensor unit 21 when the denture 2 is stationary in a first direction, second acceleration data in three-axis directions measured by the acceleration sensor of the sensor unit 21 when the denture 2 is stationary in a second direction, and third acceleration data in three-axis directions measured by the acceleration sensor of the sensor unit 21 when the denture 2 is stationary in a third direction.

[0113] 3. Based on the relationship between the orientations of the three dentures 2 (first direction, second direction, and third direction) and the three-axis acceleration data corresponding to each orientation (first acceleration data, second acceleration data, and third acceleration data), a calculation formula is derived for estimating the posture of the user's head based on the three-axis acceleration data corresponding to an arbitrary state of the user's head (including the lower jaw) being at rest.

[0114] The analyzing unit 42 can estimate (calculate) the user's head posture based on the three-axis acceleration data of the user's momentary still state (for example, when the upper and lower teeth are occluding) in the time series data D1 and D2 and the above calculation formula.

[0115] The analysis unit 42 is able to obtain analysis results corresponding to a given period (i.e., information that aggregates analysis results corresponding to each chewing interval CS included in the given period) by using the analysis model M1 and / or statistical methods, etc., as described above, for all chewing information included in a given period (in the present embodiment, the meal time for one meal).

[0116] [Proposal Module]

[0117] Next, the generator 44 inputs part or all of the analysis results corresponding to the given period into the proposal model M2 through the proposal module 44a. Thus, the generator 44 obtains the information output from the proposal model M2 as the proposal information. In addition, the generator 44 may analyze the analysis results corresponding to the given period using a statistical method or the like instead of using the proposal model M2, thereby obtaining information such as the amount of a meal, nutritional balance, number of chewing times, meal time, and bite force, and generates the proposal information based on the obtained information and the pre-defined proposal generation rule. For example, the generator 44 may predetermine a preferred range for the amount of a meal, and when the amount of a meal (estimated value) obtained from the above analysis result is outside the range, generate proposal information that prompts the amount of a meal to converge within the range. For example, when the amount of a meal (estimated value) is less than the range, the generator 44 may extract text data such as "Please increase the amount of the meal slightly." from the pre-prepared template information, and generate the text data as the proposal information. The above-mentioned proposal generation rule is information that predetermines such a rule. In addition, the proposal information may also be voice data. For example, when a sound generator such as a speaker is provided as the output unit 46 , the suggestion information may be output as sound data.

[0118] In addition to the above-mentioned information, the suggestion information generated by the generator 44 may include various information that is considered useful to the user, such as information obtained by statistical processing of the analysis results, and past user's dining history (information obtained from the analysis results). An example of the suggestion information (display items) generated by the generator 44 is shown in the following table.

[0119] [Table 1]

[0120]

[0121] [Processing flow of dietary lifestyle estimation device]

[0122] Fig.10 1 is a diagram showing an example of the processing flow of the dietary lifestyle estimation device 1. Fig.10 As shown, first, the user wears the denture 2 (step S1). Then, the user carries out daily life while wearing the denture 2 (step S2). Daily life includes mealtimes when the user ingests food. Thus, in the sensor 20 of the denture 2, the user's jaw movement information during the mealtime (the three-axis acceleration data and / or three-axis angular velocity data obtained by the sensor unit 21 in this embodiment) is acquired and saved as measurement data. The user removes the denture 2 from the user's lower jaw at a fixed time, such as before going to bed, and puts it in the storage space 34 of the base device 3 (step S3).

[0123] Next, the base device 3 acquires the measurement data from the sensor 20 of the denture 2 through the communication unit 32. In addition, the base device 3 charges the sensor 20 through the wireless charging antenna 31, and cleans the denture 2 through the cleaning unit 33 (step S4). After the user completes the processing of step S4 (for example, at a fixed time such as after getting up), the denture 2 is removed from the base device 3 (step S5) and the denture 2 is put on again (step S1). On the other hand, the base device 3 sends the measurement data acquired in step S4 when the denture 2 is placed in the storage space 34 to the user terminal 4 (step S6).

[0124] Next, in the user terminal 4, the acquisition unit 41 acquires the measurement data from the base device 3 (step S7). Next, the analysis unit 42 analyzes the measurement data (step S8). Figure 5 As shown, the analysis unit 42 determines the chewing interval CS of each mouthful through the chewing interval determination module 42a (refer to Figure 6 ), and the noise is removed for each chewing interval CS through the filtering / calculation module 42b. Thus, the chewing information of each mouthful is obtained (in this embodiment, as an example, the time series data D1 and D2). The time series data D1 and D2 are examples of time series information representing the jaw movement of a mouthful, that is, chewing information. Then, the analysis unit 42 obtains the analysis result of each mouthful from the chewing information of each mouthful through the analysis module 42c. The analysis result includes the information obtained using the analysis model M1, but may also include the information obtained without using the analysis model M1. In this embodiment, the information obtained using the analysis model M1 includes the attributes of the food that is estimated to be chewed by the user for each mouthful and the estimated value of the activity of the masticatory muscles for each mouthful. The information obtained without using the analysis model M1 is information obtained by statistical methods for the chewing information of each mouthful, such as the number of chewing times for each mouthful, the bite force for each mouthful, the posture of each mouthful (the posture of the head when eating), and other information.

[0125] Next, the generator 44 generates the above-mentioned suggestion information based on the analysis result of the analyzer 42 (step S9). Next, the output unit 46 outputs the suggestion information generated by the generator 44 (step S10). In the present embodiment, the output unit 46 outputs the display data to the display unit 47. The display unit 47 displays the display data (suggestion information) acquired from the output unit 46 (step S11). Thus, the suggestion information is presented to the user.

[0126] [Effects]

[0127] The dietary lifestyle estimation device 1 (particularly the user terminal 4 in this embodiment) includes an analysis unit 42. In such a user terminal 4, the attributes of the food estimated to be chewed by the user are obtained by using the analysis model M1 generated by machine learning. In addition, the analysis model M1 is machine-learned using training data obtained from the past meals of the user of the analysis object. Therefore, according to the user terminal 4, the analysis result (the attribute of the food estimated to be chewed by the user) can be obtained simply and with high accuracy using the analysis model M1 that has learned the characteristics of the jaw movement of the user when eating.

[0128] In addition, as described above, the analytical model M1 may also be a learned model obtained by machine learning using training data, wherein the training data includes: time series information of jaw movement corresponding to the first biting action, that is, first chewing information (in this embodiment, time series information of jaw movement corresponding to the first bite action) obtained from the user, that is, chewing information (first information); Figure 8 The data in the interval B1 in the above first information); the time series information of the jaw movement corresponding to the second biting action, that is, the second chewing information (in this embodiment, Figure 8 ); and information indicating the properties of the food chewed by the user through the jaw movement of the past bite (second information). The analysis unit 42 (analysis module 42c) may also extract information corresponding to the first chewing information and information corresponding to the second chewing information from the chewing information, and input the extracted information into the analysis model M1, thereby obtaining the properties of the food that is inferred to have been chewed by the user through the jaw movement of the past bite. In particular, in the jaw movement corresponding to the first biting action and the second biting action, features corresponding to the properties of the chewed food are likely to appear. Therefore, according to the above structure, by using the analysis model M1 that has been machine-learned using the jaw movement corresponding to the first biting action and the second biting action as feature quantities, it is possible to obtain the properties of the food that is inferred to have been chewed by the user with higher accuracy.

[0129] In addition, the user terminal 4 includes: an acquisition unit 41, which acquires jaw motion information indicating a time change of at least one of the acceleration in the three-axis direction and the angular velocity in the three-axis direction detected by the sensor provided on the denture worn on the user's lower jaw. Then, the analysis unit 42 (chewing interval determination module 42a and filtering / calculation module 42b) acquires chewing information based on the jaw motion information. According to this structure, the jaw motion information that is the basis of the chewing information can be acquired by the sensor 20 provided on the denture 2 worn on the user's lower jaw on a daily basis. In addition, by utilizing the sensor 20 provided on the denture 2 worn by the user in daily life including when eating, the jaw motion information can be appropriately acquired without placing an excessive burden on the user.

[0130] In addition, the user terminal 4 includes an acquisition unit 41, an analysis unit 42, a generation unit 44, and an output unit 46. In addition, the dietary life estimation device 1 includes a denture 2 and a user terminal 4 that analyzes the measurement data acquired by the sensor 20. In such a dietary life estimation device 1 and a user terminal 4, based on the analysis results of the jaw movement of the user during eating, suggestion information related to eating for the user is generated and output. As a result, it is possible to appropriately prompt the user and the like with a suggestion related to eating based on the analysis results. Therefore, according to the dietary life estimation device 1 and the user terminal 4, it is possible to appropriately provide useful information to the user.

[0131] In addition, in this embodiment, the analysis of the measurement data, the generation of the proposal information, and the display of the proposal information are performed locally (i.e., on the user terminal 4), so real-time feedback (prompt of the proposal information) can be provided to the user. In addition, it is not necessary to send the measurement data to a server for analyzing the measurement data (e.g., a device provided in a data center providing a cloud service, etc.), or to cause the server to perform the analysis process. Therefore, the communication load and the processing load of the above-mentioned server can be suppressed.

[0132] In addition, the measurement data (jaw movement information) includes information indicating the time change of at least one of the acceleration in the three-axis direction and the angular velocity in the three-axis direction detected by the sensor 20 provided on the denture 2 worn on the user's lower jaw. In the present embodiment, the measurement data is information indicating the time change of the acceleration in the three-axis direction (three-axis acceleration data). According to this structure, by utilizing the sensor 20 provided on the denture 2 worn by the user in daily life including when eating, it is possible to appropriately obtain information indicating the time change of the acceleration and / or angular velocity detected by the sensor 20 as measurement data without placing an excessive burden on the user. For example, in the past, the measurement of the activity of the masticatory muscles was performed by installing electrodes on the skin outside rather than in the user's mouth, but in the present embodiment, by utilizing the sensor 20 provided on the denture 2, the installation of such electrodes can be eliminated.

[0133] In addition, the measurement data is time series information of the jaw movement of the user when eating, and the analysis result of the analysis unit 42 includes attribute information indicating the attributes of the food that is estimated to be chewed by the user when eating. In this embodiment, the attributes of the food are the size, hardness and type of the food. Thus, it is possible to prompt the user with suggestion information corresponding to the attributes of the food that is estimated to be ingested by the user.

[0134] In addition, the analysis unit 42 extracts the time series information representing the jaw movement of a bite, that is, the chewing information (in this embodiment, the time series data D1 and D2). Then, the analysis unit 42 inputs the chewing information to the analysis model M1 generated by machine learning, and obtains the attributes of the food that is estimated to be chewed by the user through the jaw movement of a bite as attribute information output from the analysis model M1. The analysis model M1 is a learned model obtained by machine learning using training data, and the training data includes the time series information representing the jaw movement of a past bite obtained for the user, that is, the chewing information (first information), and the information representing the attributes of the food chewed by the user through the jaw movement of the past bite (second information). According to the above structure, using the analysis model M1 that has learned the characteristics of the jaw movement of the user when eating, it is possible to simply and accurately obtain the analysis result (estimated to be the attribute of the food chewed by the user). In addition, as an example in this embodiment, the analysis model M1 is configured to output an estimated value of the amount of masticatory muscle activity for a bite in addition to the attributes of the above-mentioned food. That is, the analytical model M1 is a learned model obtained by machine learning using the training data that also includes the above-mentioned past chewing muscle activity amount. That is, the output object of the analytical model M1 and the acquisition object of the analytical unit 42 may be only the attribute information of the food, or may also include the attribute information of the food and the above-mentioned estimated value of the chewing muscle activity amount for one bite. In the case where the analytical unit 42 also acquires the estimated value of the chewing muscle activity amount for one bite, the training data is configured to also include the chewing muscle activity amount for one bite as the second information.

[0135] Furthermore, the user terminal 4 includes a display unit 47 for displaying the suggestion information output by the output unit 46. By displaying the suggestion information (for example, part or all of the information corresponding to the display items shown in Table 1), the suggestion content corresponding to the analysis result can be appropriately presented to the user.

[0136] In addition, the dietary lifestyle estimation device 1 includes a base device 3 for storing the denture 2 removed from the user's lower jaw. In addition, the base device 3 includes a communication unit 32 that acquires measurement data from the sensor 20 by communicating with the sensor 20, and transmits the measurement data to the user terminal 4 by communicating with the user terminal 4. According to the above structure, the measurement data can be automatically transmitted from the base device 3 to the user terminal 4 only by the user placing the denture 2 on the base device 3. As a result, the user's convenience can be improved in acquiring (extracting) the measurement data from the sensor 20 provided on the denture 2.

[0137] The base device 3 also has a wireless charging antenna 31 for charging the sensor 20. According to the above configuration, charging by the wireless charging antenna 31 can be performed together with communication by the communication unit 32. Thus, the power of the sensor consumed in the communication operation of the communication unit 32 can be appropriately supplemented.

[0138] The base device 3 also includes a cleaning unit 33 for cleaning the denture 2. According to the above configuration, the user can simultaneously transmit the measurement data to the user terminal 4 and clean the denture 2 simply by placing the denture 2 on the base device 3. This can improve user convenience.

[0139] [Modifications]

[0140] As mentioned above, although one embodiment of the present invention has been described, the present invention is not limited to the above-mentioned embodiment.

[0141] For example, in the above-mentioned embodiment, the terminal (user terminal 4) of the user wearing the denture 2 is configured as a dietary lifestyle estimation device, but the dietary lifestyle estimation device may also be configured on a terminal different from the user terminal 4. For example, the above-mentioned dietary lifestyle estimation device (user terminal 4) may also be a terminal of the user's attending physician (e.g., a terminal in a hospital, etc.). In this case, the base device 3 may also be provided in the hospital. In this case, by placing the denture 2 on the base device 3 when the user visits the hospital regularly, the measurement data can be retrieved to the terminal of the attending physician. In addition, during diagnosis, the attending physician can refer to the analysis results of the measurement data and the proposal information corresponding to the analysis results on the terminal, while providing the user with suggestions related to eating, etc.

[0142] In addition, if Fig.11 As in the dietary life estimation device 1A (dietary life estimation system) involved in the modified example shown, the functions of the above-mentioned dietary life estimation device (especially the user terminal 4) can also be configured on the server. The dietary life estimation device 1A is different from the dietary life estimation device 1 in that the user terminal 4 and the server 5 are replaced with the user terminal 4A and the server 5A. Specifically, in the dietary life estimation device 1A, the server 5A has a part of the functions possessed by the user terminal 4 (the acquisition unit 41, the analysis unit 42, the analytical model storage unit 43, the generation unit 44, and the proposed model storage unit 45). The server 5A has an output unit 46A. The user terminal 4A has a display unit 47A similar to the display unit 47. In the dietary life estimation device 1A, the computing device AS is constituted by the server 5A.

[0143] When the user terminal 4A acquires measurement data from the base device 3, it does not analyze the measurement data itself but transmits the measurement data to the server 5A. The server 5A is configured to analyze the measurement data and generate proposal information based on the analysis result, similarly to the user terminal 4 described above.

[0144] The output unit 46A may also notify the user terminal 4A held by the user wearing the denture 2 of the suggestion information. Thus, the suggestion information notified from the output unit 46A is displayed on the display unit 47A of the user terminal 4A. According to the above structure, when the dietary lifestyle estimation device is configured as a server 5A on the cloud as in this modification, the suggestion information can be appropriately presented to the user.

[0145] In addition, the output unit 46A can also notify the proposed information to the terminal held by other users different from the user wearing the denture 2, that is, the terminal set as the pre-notification object. For example, the server 5A can pre-register and save the contact information (such as email addresses, etc.) of other users associated with the user for each user. In this variant, the terminal 4B and the terminal 4C are set as the notification objects. Terminal 4B is a terminal held by other users who are family members or close relatives of the user. Terminal 4C is a terminal held by other users who are the user's attending physician. Thus, the proposed information notified from the output unit 46A is displayed on the display unit 47B of the terminal 4B and the display unit 47C of the terminal 4C respectively. According to the above structure, other users can understand the user's eating status. Thus, the user's family members / close relatives, attending physicians, etc. can monitor the user's eating status.

[0146] In addition, it is also conceivable that the user wears the denture 2 for a long time. In this case, the server 5A does not perform analysis of the measurement data and generation of the proposal information, and does not notify the user's family / near relatives, attending physician, etc. of the proposal information, so the above-mentioned monitoring cannot be properly implemented. Therefore, for example, the server 5A (for example, the output unit 46A) may also notify the terminal 4B, 4C of other users who have previously established an association with the user when the state of not generating proposal information for a certain user continues for more than a predetermined period. In addition, such notification may also be performed by the base device 3 or the user terminal 4A. For example, the base device 3 may also notify the terminal 4B, 4C of other users when the state of not placing the denture 2 in the storage space 34 of the base device 3 for more than a predetermined period since the last time the denture 2 was placed in the storage space 34 of the base device 3 and communication, charging, and cleaning were performed. Similarly, the user terminal 4A may also notify the terminal 4B, 4C of other users when the state of not being able to obtain measurement data from the base device 3 continues for more than a predetermined period. Furthermore, the user terminal 4A may display notification information for urging the user to place the denture 2 on the base device 3 on the display unit 47A.

[0147] In addition, in the above-mentioned embodiment, the analytical model M1 is prepared for each user. That is, the analytical model M1 is generated only based on the training data obtained from the eating action performed by the user who is the object of analysis in the past. However, the analytical model M1 can also be a learned model obtained by machine learning using the training data, and the training data includes chewing information (first information) representing the jaw movement of the past mouthful obtained for an unspecified user (a user wearing a denture provided with a sensor in the same manner as the denture 2), and the attributes of the food chewed by the jaw movement of the unspecified user in the past mouthful (second information). That is, the training data used for learning the analytical model M1 can also be data obtained from the eating actions performed by multiple unspecified users in the past. In the case where the training data is obtained from the eating actions performed by multiple unspecified users in the past, it can also be learned that when a certain user becomes the object of analysis, the analytical result corresponding to the second information in the first information that is most suitable for the user is output. In this case, the analytical model M1 can be universalized among multiple users. Thus, there is an advantage that the analytical model M1 does not need to be produced and managed for each user. In addition, compared with the case where the analysis model M1 is prepared for each user as in the above-described embodiment, there is also an advantage that it is easier to collect training data for the analysis model M1.

[0148] Furthermore, if Fig.12As shown, the analytical model M1 may also be a learned model obtained by machine learning using training data that also includes a user profile representing attributes of an unspecified user. In this case, the analyzing unit 42 is configured to also input a user profile (profile information) representing the attributes of the user (the user to be analyzed) into the analytical model M1. Here, the attributes of the user are, for example, gender, age (or generation), health status, etc. The jaw movement of the user when eating can be considered to be different depending on such user attributes. Therefore, according to the above structure, by including the user's attributes in the feature amount, the attributes of the food presumed to be chewed by the user can be obtained with higher accuracy.

[0149] In addition, the analysis result may also include an opening amount indicating the size of the user's mouth opening when ingesting food. Such an opening amount can be determined based on the chewing information ( Fig. 9 The waveform of the time series data D3) can be used to grasp the situation. Furthermore, the generation unit 44 can also generate proposal information containing information related to whether the diagnosis of mandibular arthritis is needed based on the amount of mouth opening. For example, the generation unit 44 obtains the maximum value (or average value, etc.) of the amount of mouth opening of the user in a given period (for example, a daily amount) based on the analysis result corresponding to the given period, and compares the maximum value with a predetermined threshold. Furthermore, when the maximum value is less than the threshold, the generation unit 44 can determine that there is a suspicion of mandibular arthritis and generate proposal information urging the diagnosis of mandibular arthritis. According to such a structure, it is possible to urge early diagnosis to users who are suspected of mandibular arthritis based on the size of the user's mouth opening when eating.

[0150] In addition, the generating unit 44 may also store information related to the user's usual meal size in advance, calculate the total meal size of the user in the given period based on the analysis result corresponding to the given period (mainly the estimated size of each bite of food), compare the total meal size with the above-mentioned usual meal size, and generate suggestion information corresponding to the comparison result. For example, the generating unit 44 may determine that there is a suspicion of gastric or intestinal disease when the above-mentioned total meal size is less than the above-mentioned usual meal size by more than a predetermined threshold, and generate suggestion information urging to check the condition of the stomach or intestine. Alternatively, the generating unit 44 may determine that the user eats too much when the above-mentioned total meal size is more than the above-mentioned usual meal size by more than a predetermined threshold, and generate suggestion information urging to reduce the meal size.

[0151] In addition, for example, in the case where the user wears a complete denture, a sensor similar to the sensor 20 of the denture 2 worn on the lower jaw may be provided on the denture worn on the upper jaw. Furthermore, the measurement data acquired by the acquisition unit 41 may include information indicating the time change of at least one of the acceleration in the three-axis direction and the angular velocity in the three-axis direction detected by the sensor provided on the denture worn on the upper jaw of the user (hereinafter referred to as "upper jaw measurement data") together with the measurement data detected by the sensor 20 (hereinafter referred to as "lower jaw measurement data"). For example, when the user is riding in a vehicle such as a car or a train, the vibration component from the vehicle is detected by the sensor 20 and may be mistakenly detected as the jaw movement of the user when eating. In addition, when the user eats a meal while riding in a vehicle, the vibration component from the vehicle may be mixed into the jaw movement of the user when eating as noise. On the other hand, according to the above structure, the analysis unit 42 can perform analysis using not only the detection results of the sensor 20 provided on the denture 2 worn on the user's lower jaw, but also the detection results of the sensor provided on the denture worn on the user's upper jaw. Specifically, the chewing interval determination module 42a can obtain (extract) the motion component representing only the user's jaw movement in the lower jaw measurement data as chewing information by using not only the lower jaw measurement data but also the upper jaw measurement data. For example, the chewing interval determination module 42a can eliminate the vibration component from the vehicle (i.e., the component contained in both the upper jaw sensor and the lower jaw sensor) by obtaining the relative value of the detection result of the sensor of the lower jaw (i.e., the lower jaw measurement data) relative to the detection result of the sensor of the upper jaw (i.e., the upper jaw measurement data). As a result, the above-mentioned problem can be eliminated.

[0152] In addition, in the above-mentioned embodiment, the user terminal 4 or the server 5A as a dietary life estimation device of one embodiment has an analysis function of obtaining the attributes of the food estimated to be chewed by the user as an analysis result by analyzing the measurement data, and has a suggestion function of making a suggestion based on the analysis result. However, in the user terminal 4 or the server 5A as a dietary life estimation device, the above-mentioned suggestion function (i.e., the generation unit 44, the suggestion model storage unit 45, the output unit 46, and the display unit 47, etc.) is not essential. That is, the part that performs the suggestion function described in the above-mentioned embodiment can also be cut out from the user terminal 4 or the server 5A to another device.

[0153] In addition, if Fig.13 As in the dietary lifestyle estimation device 1B (dietary lifestyle estimation system) according to the modified example shown in FIG. Figure 1The user terminal 4 shown in the figure is a user device 4D to which the function of the server 5 is added, and the server is omitted. The user device 4D may be independently installed in the user's home, etc. Alternatively, the user device 4D may communicate with the user via any communication network such as a telephone line and an Internet line. Fig.11 The same terminal 4B as shown in the figure is the terminal 4E, Fig.11 The terminal 4C shown in the figure is similar to the terminal 4F, etc., which notifies information (for example, information output by the output unit 46). The functions of the user device 4D, the terminal 4E, and the terminal 4F are the same as the functions of the user terminal 4, the server 5, the terminal 4B, and the terminal 4C mentioned above, so detailed descriptions are omitted. When the user device 4D is connected to the terminal 4E and the terminal 4F, it is possible to appropriately notify the user's relatives (for example, family members, attending physicians, etc.) of suggestion information and warnings. In particular, when the user's relatives live separately from the user, early detection and early response to adverse conditions caused by the user can be achieved. In the dietary life estimation device 1B, the user device 4D constitutes the computing device AS.

[0154] The embodiments described in the present invention can also be expressed as follows.

[0155] [Invention method 1]

[0156] A system for performing dietary estimation comprises a circuit,

[0157] The circuit performs the following analysis: obtaining chewing information, which is time series information representing the jaw movement of a bite amount when the user eats, inputting the chewing information into an analysis model generated by machine learning, and obtaining attributes of food that are estimated to be chewed by the user through the jaw movement of the bite amount, output from the analysis model,

[0158] The analytical model is a learned model obtained by machine learning using training data, wherein the training data includes first information, time series information representing jaw movements of past bites obtained for the user, and second information representing attributes of food chewed by the user through jaw movements of the past bites.

[0159] [Invention method 2]

[0160] A system for performing dietary estimation comprises a circuit,

[0161] The circuit performs the following analysis: obtaining chewing information, which is time series information representing the jaw movement of a bite amount when the user eats, inputting the chewing information into an analysis model generated by machine learning, and obtaining attributes of food that are estimated to be chewed by the user through the jaw movement of the bite amount, output from the analysis model,

[0162] The analytical model is a learned model obtained by machine learning using training data, wherein the training data includes first information, time series information representing jaw movements of past bites obtained for an unspecified user, and second information representing attributes of food chewed by the unspecified user through the jaw movements of the past bites.

[0163] [Invention mode 3]

[0164] Based on the system for performing dietary estimation according to the second embodiment of the invention,

[0165] The circuit further inputs profile information representing attributes of the user into the analysis model when performing the analysis.

[0166] The analytical model is a learned model obtained by performing machine learning using the training data that also includes the profile information of the unspecified user.

[0167] [Invention mode 4]

[0168] In the system for performing dietary estimation according to the first aspect of the invention,

[0169] The analytical model is a learned model obtained by machine learning using training data, the training data including data corresponding to first chewing information, which is time series information representing jaw movement corresponding to the first biting action in the first information, data corresponding to second chewing information, which is time series information representing jaw movement corresponding to the second biting action in the first information, and the second information,

[0170] When performing the analysis, the circuit extracts the first chewing information and the second chewing information from the chewing information and inputs the extracted information into the analysis model, thereby acquiring properties of food estimated to have been chewed by the user through the jaw movement of the bite.

[0171] [Invention mode 5]

[0172] In the system for performing dietary estimation according to the first aspect of the invention,

[0173] The circuit acquires jaw movement information detected by a sensor provided on a denture worn on a lower jaw of the user, and acquires the chewing information based on the jaw movement information during the analysis.

[0174] [Invention mode 6]

[0175] Based on the system for performing dietary estimation according to the fifth embodiment of the invention,

[0176] The jaw movement information includes information indicating a temporal change in at least one of acceleration in three-axis directions and angular velocity in three-axis directions detected by the sensor.

[0177] [Invention mode 7]

[0178] Based on the system for performing dietary estimation according to the sixth embodiment of the invention,

[0179] The jaw movement information further includes information indicating a temporal change in at least one of acceleration in three-axis directions and angular velocity in three-axis directions detected by a sensor provided on the artificial tooth worn on the upper jaw of the user.

[0180] [Invention mode 8]

[0181] In the system for performing dietary estimation according to the first embodiment of the invention,

[0182] The attributes of the food include at least one of size, hardness and type of the food.

[0183] [Invention method 9]

[0184] Based on the system for performing dietary estimation according to the fifth embodiment of the invention,

[0185] The system further comprises a base for storing the denture removed from the lower jaw of the user,

[0186] The base acquires the jaw movement information from the sensor by communicating with the sensor.

[0187] [Invention mode 10]

[0188] Based on the system for performing dietary estimation according to the ninth aspect of the invention,

[0189] The base also has a charger for charging the sensor.

[0190] [Invention mode 11]

[0191] Based on the system for performing dietary lifestyle estimation according to the ninth aspect of the invention,

[0192] The base also has a cleaning device for cleaning the denture.

[0193] [Invention mode 12]

[0194] A device for performing dietary estimation comprises a circuit,

[0195] The circuit performs the following analysis: obtaining chewing information, which is time series information representing the jaw movement of a bite amount when the user eats, inputting the chewing information into an analysis model generated by machine learning, and obtaining attributes of food that are estimated to be chewed by the user through the jaw movement of the bite amount, output from the analysis model,

[0196] The analytical model is a learned model obtained by machine learning using training data, wherein the training data includes first information, time series information representing jaw movements of past bites obtained for the user, and second information representing attributes of food chewed by the user through jaw movements of the past bites.

[0197] [Invention mode 13]

[0198] A device for performing dietary estimation comprises a circuit,

[0199] The circuit performs the following analysis: obtaining chewing information, which is time series information representing the jaw movement of a bite amount when the user eats, inputting the chewing information into an analysis model generated by machine learning, and obtaining attributes of food that are estimated to be chewed by the user through the jaw movement of the bite amount, output from the analysis model,

[0200] The analytical model is a learned model obtained by machine learning using training data, wherein the training data includes first information, time series information representing jaw movements of past bites obtained for an unspecified user, and second information representing attributes of food chewed by the unspecified user through the jaw movements of the past bites.

[0201] [Invention mode 14]

[0202] A method for estimating dietary habits, comprising the steps of: obtaining time series information representing a bite-sized jaw movement of a user when ingesting food, i.e., chewing information; inputting the chewing information into an analytical model generated by machine learning; and performing analysis to obtain attributes of food estimated to have been chewed by the user through the bite-sized jaw movement output from the analytical model,

[0203] The analytical model is a learned model obtained by machine learning using training data, wherein the training data includes first information, time series information representing jaw movements of past bites obtained for the user, and second information representing attributes of food chewed by the user through jaw movements of the past bites.

[0204] [Invention mode 15]

[0205] A method for estimating dietary habits, comprising the steps of: obtaining time series information representing a user's jaw movement for a bite-sized amount when eating, i.e., chewing information; inputting the chewing information into an analytical model generated by machine learning; and performing analysis to obtain properties of food output from the analytical model that is estimated to be chewed by the user through the jaw movement for the bite-sized amount,

[0206] The analytical model is a learned model obtained by machine learning using training data, wherein the training data includes first information, time series information representing jaw movements of past bites obtained for an unspecified user, and second information representing attributes of food chewed by the unspecified user through the jaw movements of the past bites.

[0207] (Number Description)

[0208] 1, 1A, 1B...dietary life estimation device, 2...denture, 4...user terminal, 5A...server, 20...sensor, 41...acquisition unit, 42...analysis unit, 42a...chewing interval determination module, 42b...filtering / calculation module, 42c...analysis module, M1...analysis model.

Claims

1. A dietary lifestyle estimation device, comprising an analysis unit, the analysis unit acquiring chewing information, which is time series information representing jaw movements of a bite-sized portion when a user ingests food, inputting the chewing information into an analysis model generated by machine learning, and acquiring attributes of food that are estimated to have been chewed by the user through the jaw movements of the bite-sized portion, output from the analysis model; The analytical model is a learned model obtained by machine learning using training data, the training data including time series information of jaw movement representing past bites obtained for the user, namely, data corresponding to first chewing information in first information, data corresponding to second chewing information, and second information representing properties of food chewed by the user through jaw movement of the past bites, the first chewing information being time series information of jaw movement corresponding to the first biting action in the first information, the second chewing information being time series information of jaw movement corresponding to the second biting action in the first information, The analysis unit extracts the first chewing information and the second chewing information from the chewing information and inputs the extracted information into the analysis model, thereby acquiring the attribute of the food estimated to have been chewed by the user through the jaw movement of the bite.

2. A dietary lifestyle estimation device comprising an analysis unit that obtains chewing information, which is time series information representing jaw movements of a bite-sized portion when a user ingests food, inputs the chewing information into an analysis model generated by machine learning, and obtains attributes of food that are estimated to have been chewed by the user through the jaw movements of the bite-sized portion, output from the analysis model. The analytical model is a learned model obtained by machine learning using training data, wherein the training data includes time series information of jaw movements representing past bites obtained for an unspecified user, namely, data corresponding to first chewing information in first information, data corresponding to second chewing information, and second information representing properties of food chewed by the unspecified user through jaw movements of the past bites, wherein the first chewing information is time series information of jaw movements corresponding to the first biting action in the first information, and the second chewing information is time series information of jaw movements corresponding to the second biting action in the first information. The analysis unit extracts the first chewing information and the second chewing information from the chewing information and inputs the extracted information into the analysis model, thereby acquiring the attribute of the food estimated to have been chewed by the user through the jaw movement of the bite.

3. The dietary lifestyle estimation device according to claim 2, wherein: The analysis unit further inputs profile information representing attributes of the user into the analysis model, The analytical model is a learned model obtained by performing machine learning using the training data that also includes the profile information of the unspecified user.

4. The dietary lifestyle estimation device according to any one of claims 1 to 3, wherein: The dietary lifestyle estimation device further includes an acquisition unit that acquires jaw movement information detected by a sensor provided on a denture worn on a lower jaw of the user. The analyzing unit acquires the mastication information based on the jaw movement information.

5. The dietary lifestyle estimation device according to claim 4, wherein: The jaw movement information includes information indicating a temporal change in at least one of acceleration in three-axis directions and angular velocity in three-axis directions detected by the sensor.

6. The dietary lifestyle estimation device according to claim 5, wherein: The jaw movement information further includes information indicating a temporal change in at least one of acceleration in three-axis directions and angular velocity in three-axis directions detected by a sensor provided on the artificial tooth worn on the upper jaw of the user.

7. The dietary lifestyle estimation device according to any one of claims 1 to 3, wherein: The attributes of the food include at least one of size, hardness and type of the food.

8. The dietary lifestyle estimation device according to claim 4, wherein: The dietary lifestyle estimation device further includes a base device for storing the denture removed from the lower jaw of the user. The base device acquires the jaw movement information from the sensor by communicating with the sensor.

9. The dietary lifestyle estimation device according to claim 8, wherein: The base device further includes a charging unit for charging the sensor.

10. The dietary lifestyle estimation device according to claim 8, wherein: The base device also has a cleaning portion for cleaning the denture.

Images