Massage system

Abstract

This provides new technology for generating massage courses according to user requests or circumstances. [Solution] The disclosed massage system inputs input data relating to the user of the massage machine into a large-scale language model that generates massage motion data for causing the massage machine to perform a massage operation, outputs massage motion data corresponding to the input data from the large-scale language model, and causes the massage machine to perform a massage operation based on the massage motion data output from the large-scale language model.

Description

Technical Field

[0001] This disclosure relates to a massage system.

Background Art

[0002] A massage course for executing a plurality of preset massage operations in a predetermined order is set in a massage machine.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Patent Document 2

Summary of the Invention

[0004] With only the preset massage course in the massage machine, it is impossible to meet various demands of users for the massage by the massage machine or the user's situation.

[0005] Therefore, a new technology for generating a massage course according to the user's demands or the user's situation is desired.

[0006] One aspect of this disclosure is a system. The disclosed system inputs input data regarding a user of a massage machine into a large language model that generates massage operation data for causing the massage machine to perform a massage operation, causes the massage operation data corresponding to the input data to be output from the large language model, and can be a massage system that causes the massage machine to perform a massage operation based on the massage operation data output from the large language model.

[0007] Further details will be described as embodiments below.

Brief Description of the Drawings

[0008] [Figure 1] Figure 1 is a schematic diagram of the massage system. [Figure 2] Figure 2 is a block diagram of the massage machine. [Figure 3] Figure 3 is a schematic side view of the massage machine. [Figure 4] Figure 4 is a block diagram of the generation system. [Figure 5] Figure 5 is a flowchart for generating massage courses. [Figure 6] Figure 6 is a block diagram of the generation system. [Figure 7] Figure 7 is a flowchart for generating massage courses. [Figure 8] Figure 8 is an explanatory diagram of the massage course. [Figure 9] Figure 9 shows the operation of the massage course. [Modes for carrying out the invention]

[0009] <1. Overview of the massage system>

[0010] (1) The massage system according to the embodiment inputs input data relating to the user of the massage machine into a large-scale language model that generates massage motion data for causing the massage machine to perform a massage operation, outputs massage motion data corresponding to the input data from the large-scale language model, and causes the massage machine to perform a massage operation based on the massage motion data output from the large-scale language model. With this massage system, massage motion data corresponding to input data relating to the user can be generated by the large-scale language model.

[0011] (2) The large-scale language model may be configured to generate massage operation data corresponding to the functions of the massage section of the massage machine, based on the model information of the massage machine. In this case, the large-scale language model can generate massage operation data corresponding to the model of the massage machine.

[0012] (3) The large-scale language model may comprise a first large-scale language model and a second large-scale language model. The first large-scale language model may be configured to generate intermediate data from the input data, which is then given as input to the second large-scale language model. The second large-scale language model may be configured to generate the massage motion data from the intermediate data.

[0013] (4) The second large-scale language model may be configured to generate massage operation data corresponding to the functions of the massage section of a massage machine based on the model information of the massage machine. In this case, the second large-scale language model can generate massage operation data corresponding to the model of the massage machine. On the other hand, the first large-scale language model is preferable as it does not need to correspond to the model of the massage machine.

[0014] (5) The intermediate data may include body part data indicating the body part of the user to be massaged. The first large-scale language model may generate the intermediate data including the body part data from the input data. The second large-scale language model may generate the massage motion data from the intermediate data including the body part data.

[0015] (6) The second large-scale language model may be configured to generate massage operation data that causes the massage unit of the massage machine corresponding to the model information to perform a massage operation on the body part indicated by the body part data, based on the model information of the massage machine and the body part data. In this case, the massage operation data that causes the massage operation on the body part indicated by the body part data can be generated according to the model of the massage machine.

[0016] (7) The massage system can determine the suitability of the massage operation data output from the large language model. Since the large language model may generate inappropriate massage operation data, determining the suitability enables handling of inappropriate massage operation data.

[0017] (8) The massage system determines the suitability of the massage operation data output from the large language model, and if it is determined to be inappropriate, can cause the large language model to output corrected massage operation data. Since the large language model may generate inappropriate massage operation data, it can be corrected if it is inappropriate.

[0018] (9) The input data may include age-related data regarding the user's age. The massage operation data output from the large language model can generate at least one of part data indicating parts of the user that should not be massaged and adjustment data for adjusting the massage operation, based on the age-related data.

[0019] <2. Example of a Massage System>

[0020] Hereinafter, embodiments will be described in more detail while referring to the drawings.

[0021] FIG. 1 shows a massage system 500 according to an embodiment. The massage system 500 may include a generation system 300 that generates operation data indicating operations performed by the massage machine 10. The generation system 300 can generate massage course data (massage operation data) for massage operations as the operation data. The massage machine 10 can acquire the massage operation data from the generation system 300 and execute a massage operation based on the massage operation data.

[0022] The massage machine 10 comprises a main unit 11. The main unit 11 is configured, for example, in the form of a chair. The main unit 11 supports the body of the person receiving treatment 1, who is the user of the massage machine 10. The main unit 11 comprises, for example, a seat 13, a backrest 15, and a footrest 17. The seat 13 can support the buttocks of the person receiving treatment 1.

[0023] The backrest 15 shown in Figure 1 is reclinably mounted at the rear of the seat 13. The backrest 15 can support the back of the patient 1. The backrest 15 can also support the head of the patient 1. The recline of the backrest 15 can be driven by a reclining actuator 21 (see Figure 2). The reclining actuator 21 is, for example, an electric cylinder or a pneumatic cylinder.

[0024] The footrest 17 shown in Figure 1 is mounted on the front of the seat 13 so as to be able to swing up and down. The footrest 17 can support the legs of the person receiving treatment 1. The swinging motion of the footrest 17 can be driven by a swing actuator 23. The swing actuator 23 is, for example, an electric cylinder or a pneumatic cylinder.

[0025] The massage machine 10 includes a massage unit that performs actions on the person being treated 1. For example, the backrest 15 may have a first massage unit 100 inside. The first massage unit 100 mainly massages the upper body of the person being treated 1, such as the back or neck. The first massage unit 100 may also be configured to massage the lower body.

[0026] Figure 3 shows an example of the schematic configuration of the first massage unit 100. The first massage unit 100 is installed inside the backrest 15 so as to be movable in the longitudinal direction Y of the body of the person being treated 1. The first massage unit 100 shown in Figure 3 is installed so as to be movable in the longitudinal direction of the backrest 15. The first massage unit 100 may also be movable within the range of the seat 13 in order to perform a massage action on the lower body.

[0027] The massage machine 10 shown in Figure 3 includes a moving mechanism 160 for moving the first massage unit 100. The moving mechanism 160 is composed of, for example, a linear actuator 162 for moving the first massage unit 100 within the backrest 15. The linear actuator 162 is composed of, for example, a ball screw actuator and is driven by a lifting motor 161. The first massage unit 100 can move in the longitudinal direction Y of the body by the rotation of the lifting motor 161. The first massage unit 100 can perform a massage operation on the person being treated 1 while moving in the longitudinal direction Y of the body, or it can stop at any position in the longitudinal direction Y of the body and perform a massage operation on the person being treated 1.

[0028] The first massage unit 100 may include a treatment element 150 that applies a massage motion to the person being treated 1. In the first massage unit 100 shown in Figure 3, the treatment element 150 is provided to be movable in the forward / backward direction X. That is, in the first massage unit 100 shown in Figure 3, the treatment element 150 is provided to be able to move forward and backward relative to the person being treated 1. Because the treatment element 150 is movable in direction X, the strength of the massage motion can be adjusted. That is, when the treatment element 150 is moved forward, which is the direction towards the person being treated 1, a strong massage motion is obtained, and when the treatment element 150 is moved backward, which is the direction away from the person being treated 1, a weak massage motion is obtained.

[0029] To allow the treatment element 150 to move freely in direction X, the first massage unit 100 shown in Figure 3 may include a base unit 120 attached to a moving mechanism 160 and a reciprocating unit 130 supported by the base unit 120. The base unit 120 is provided to move freely in direction Y by the moving mechanism 160. The reciprocating unit 130 is provided to move freely in direction X relative to the base unit 120.

[0030] The forward / backward unit 130 moves in direction X by, for example, a strength adjustment motor 121 provided on the base unit 120 or the forward / backward unit 130. The forward / backward unit 130 may also move in direction X by an air cell provided between the base unit 120 and the forward / backward unit 130. The air cell is configured to expand and contract by supplying and exhausting air. When the air cell expands, the forward / backward unit 130 moves forward, and when the air cell contracts, the forward / backward unit 130 can move backward.

[0031] The first massage unit 100 includes a massage mechanism that generates a massage motion on the therapist 150 to be applied to the person being treated 1. The massage mechanism converts the rotation of a kneading motor 131 and a tapping motor 132 (see Figure 2) into a massage motion using a mechanical mechanism with gears, and transmits the massage motion to the therapist 150 via the arm 140. The massage mechanism may, for example, perform the massage motion by the expansion and contraction of an air cell, or convert the expansion and contraction of an air cell into a massage motion. In Figures 2 and 3, a kneading motor 131 for kneading and a tapping motor 132 for tapping are provided, but both kneading and tapping may be performed by a single motor. For example, the massage mechanism may be configured so that kneading is performed when the motor rotates in one direction, and tapping is performed when it rotates in the opposite direction.

[0032] The massage mechanism is configured, for example, to cause the treatment element 150 to perform kneading and tapping actions. The kneading action is, for example, an action in which a pair of treatment elements 150 repeatedly move closer to and away from each other. The tapping action is, for example, an action in which a pair of treatment elements 150 repeatedly move back and forth alternately against the person being treated 1. The massage mechanism according to the embodiment is configured to be able to perform both kneading and tapping actions, and can selectively perform either kneading or tapping actions. The massage mechanism may also be configured to perform a kneading-tapping action, where kneading and tapping actions are performed simultaneously. Furthermore, by moving the first massage unit 100 in direction Y while applying the treatment element 150 to the person being treated 1, a rolling massage action can be applied to the person being treated 1. In addition, by pressing the treatment element 150 against the person being treated 1 without moving the motors 131 and 132, acupressure action can be applied. Thus, the first massage unit 100 according to this embodiment can perform multiple types of massage operations.

[0033] As shown in Figure 1, the massage machine 10 may include a second massage unit 200 as a massage unit other than the first massage unit 100. The second massage unit 200 includes, for example, an air cell 201, which performs a massage on the person receiving treatment 1 through the expansion and contraction of the air cell 201. The air cell 201 is positioned, for example, on the footrest 17 and massages the legs of the person receiving treatment 1. The air cell 201 may also be positioned in other locations on the main body 11 of the device, such as the seat 13 or the backrest 15.

[0034] As shown in Figure 2, the massage machine 10 may be equipped with an air circuit 203 for supplying and exhausting air to and from the air cell 201. The air circuit 203 may include an air pressure source such as a pump, a hose for supplying air, a solenoid valve for switching the airflow, and so on.

[0035] As shown in Figure 2, the massage machine 10 may be equipped with a controller 30. The controller 30 can control the operation of the first massage unit 100 and the second massage unit 200. The controller 30 can also control the reclining of the backrest 15 and the swinging of the footrest 17. To control these operations, the controller 30 is connected to various motors 131, 132, 161, and 121. The controller 30 is also connected to other actuators 21, 23 and an air circuit 203.

[0036] The controller 30 is connected to various sensors 40. The massage machine 10 may be equipped with various sensors for controlling its operation. For example, the massage machine 10 may be equipped with a position sensor that detects the position of the first massage unit 100 in direction Y. The massage machine 10 may also be equipped with a position sensor that detects the position of the reciprocating unit 130 in direction X. The massage machine 10 may also be equipped with a sensor that detects the pressing force of the treatment element 150 on the person being treated 1.

[0037] As shown in Figure 2, the controller 30 may be composed of a computer comprising a processor 30A and a storage device 30B. The processor 30A is connected to the storage device 30B. The processor 30A can read and execute a computer program 30D stored in the storage device 30B. The computer program 30D may include program code that indicates commands for controlling the massage machine 10.

[0038] The storage device 30B can store one or more massage course data. Massage course data (massage action data) is data that sets a series of actions for a massage course. Multiple massage course data may include, for example, a first massage course data 31 and a second massage course data 32. The processor 30A reads one massage course data selected by the user 1 from among the multiple massage course data 31, 32 from the storage device 30B and executes it. That is, the massage course data (massage action data) is a computer program that includes instructions to the controller 30, and the controller 30 that has executed the massage course data controls the massage unit to make the massage unit perform a predetermined massage action.

[0039] The storage device 30B stores pre-set massage course data 31 and 32, and can also store massage course data received from an external source such as the generation system 300 shown in Figure 1, or massage course data generated by the controller 30.

[0040] The series of actions set in the massage course data constitutes a sequence for the massage course. The sequence for the massage course is configured so that the controller 30 executes a set of pre-set actions in a predetermined order.

[0041] The controller 30 executes a sequence for a massage course according to the massage course data, which is a computer program, and controls the operation of the massage units 100, 200, etc., according to the operations set in the sequence.

[0042] The controller 30 includes a communication unit 30C for communicating with an external device of the massage machine 10. The external device is, for example, a generation system 300 (see Figure 1), or a user device 600 owned by the user, who is the person receiving treatment 1. The user device 600 is, for example, a smartphone, tablet, or wearable device owned by the user.

[0043] The controller 30 can communicate with external devices wirelessly or via wired connection through the communication unit 30C, and can transmit or receive data. For example, the controller 30 can receive massage course data generated by the generation system 300. The controller 30 can also receive user data such as the user's vital data and other health data from the user device 600. The controller 30 stores the received data in the storage device 30B.

[0044] The massage machine 10 may be equipped with an operating device 50 for the person receiving treatment 1 to operate the massage machine 10. The operating device 50 is connected to a controller 30. The connection between the operating device 50 and the controller 30 may be a wired connection or a wireless connection.

[0045] The operating device 50 may include input / output devices. These input / output devices may include, for example, an output device such as a display 51, and an input device such as various operation buttons and a microphone 52. The display 51 is preferably a touch panel display. The display 51 and microphone 52, etc., may be used by the controller 30 to obtain data necessary for generating massage courses from the user.

[0046] The operation buttons may include, for example, buttons for selecting pre-set massage courses. The massage courses are assigned course names such as "Full Body," "Foot Reflexology," "Neck and Shoulders," "Posture Stretching," "Body Defense," and "Body Activation," and these names are attached to the selection buttons. The massage courses may be displayed on the display 51 and selected on the display 51.

[0047] When a massage course is selected by the patient 1, the controller 30 reads the massage course data 31 and 32 corresponding to the selected massage course, and starts executing the sequence for the massage course according to the read massage course data 31 and 32. The controller 30 controls the massage units 100, 200, etc., according to the operations set in the sequence. The massage course data 31 and 32 have a set of appropriate operations corresponding to their course names.

[0048] The massage machine 10 can receive and execute massage course data 31 and 32 pre-set in the controller 30, as well as massage course data generated by the generation system 300. Therefore, the user, who is the person receiving treatment 1, can use massage courses other than those pre-set in the massage machine 10. The generated massage course data is stored in the storage device 30B of the controller 30 and can be used for subsequent execution.

[0049] As shown in Figure 1, the generation system 300 is configured, for example, as a device connected to the massage machine 10 via a network 400 such as the Internet. The generation system 300 may, for example, be configured as a server computer connected to the Internet. The generation system 300 does not need to be provided separately from the massage machine 10; it may be provided on the massage machine 10. In other words, the massage system 500 may be provided integrally with the massage machine 10.

[0050] The generation system 300 shown in Figure 1 may be comprised of a computer comprising one or more processors 300A and a storage device 300B. The generation system 300 may also include a communication unit 300C for communicating with the outside world.

[0051] The processor 300A is connected to the storage device 300B. The processor 300A can read and execute a computer program stored in the storage device 300B. The computer program may include program code for generating massage course data (massage motion data) that can be executed by the massage machine 10. The computer program, when executed by the processor 300A, produces the functions of each element 310, 311, 312, 320 of the generation system 300 (see Figures 4 and 6).

[0052] The processor 300A receives data from the massage machine 10 via the network 400 using the communication unit 300C, and executes a process to generate massage course data based on the received data. The processor 300A transmits the generated massage course data from the communication unit 300C, and the transmitted massage course data can be received by the massage machine 10 via the network 400.

[0053] Figure 4 shows an example of the configuration of the generation system 300. As shown in Figure 4, the generation system 300 includes a massage course data generator 310 (massage motion data generator). As mentioned above, the massage course data is a computer program that includes commands to the controller 30, and the controller 30, having executed the massage course data, controls the massage unit to make each massage unit perform a predetermined massage operation.

[0054] The generator 310 generates massage courses that can be executed by the massage machine 10, which is provided with massage course data. The generator 310 can generate massage course data that is compatible with the model of the massage machine 10, which is provided with massage course data.

[0055] Here, the massage machine 10 may differ in the type, function, and number of massage parts depending on the model, and the controller 30 may also have different specifications. Therefore, if the massage course data is not created in a format and specifications appropriate to the model, the massage machine 10 will not be able to execute that massage course data. However, the generator 310 of this embodiment can generate massage course data that can be executed by the massage machine 10 according to the model of the massage machine 10, so it can provide executable massage course data for multiple massage machines 10 of different models.

[0056] The generator 310 generates massage course data using a large language model 310A (LLM). The large language model 310A may be built within the generation system 300 or outside the generation system 300.

[0057] The large-scale language model 310A is a type of generative AI that is trained using a large amount of data. It can understand input data such as text and generate and output responses in natural language. In addition to generating natural language, the large-scale language model 310A can also generate program code. In other words, the large-scale language model 310A has code generation capabilities.

[0058] The generator 310 of this embodiment generates massage course data consisting of program code by utilizing the code generation function of the large-scale language model 310A. In order for the large-scale language model 310A to generate massage course data, the large-scale language model 310A is pre-trained (machine learning) with data that represents knowledge of the programming language executed by the processor 30A of the controller 30 of the massage machine 10. Therefore, the large-scale language model 310A can generate a program (such as massage course data) that can be executed by the controller 30 of the massage machine 10.

[0059] In a typical large-scale language model, program code can be generated by a code generation function when natural language input indicating program requirements is provided. In contrast, the large-scale language model 310A of this embodiment goes further, understanding the user's physical condition through natural language input and generating massage course data to cause the massage machine 10 to perform massage actions corresponding to that physical condition. The large-scale language model 310A of this embodiment is trained to possess the necessary knowledge for this purpose.

[0060] For example, the large-scale language model 310A is trained with data representing knowledge for operating the massage unit and other functions controllable by the controller 30 in the massage machine 10, so that it can programmatically represent the operation of the massage machine 10, such as massage movements. For example, to control a massage unit, the values ​​of the control parameters of that massage unit (operation timing, operation duration, type of operation, intensity of operation, etc.) must be adjusted appropriately according to the type of massage unit (motor-driven massage unit, air-driven massage unit, etc.). In this case, the large-scale language model 310A can be trained with data representing knowledge about what type of massage unit the massage machine 10 has, and how to adjust which control parameters to achieve what kind of operation. Furthermore, the large-scale language model 310A is trained with a large amount of massage course data. As a result, the large-scale language model 310A can have a function (massage course data generation function) to generate massage course data (massage operation data) that realizes massage movements that can be executed in the massage machine 10.

[0061] The large-scale language model 310A is trained with data that represents knowledge about the functions of each of the various models of massage machines 10. Therefore, the large-scale language model 310A can generate massage course data that can be executed on the massage machine 10, depending on the model.

[0062] Furthermore, the large-scale language model 310A is trained with data representing knowledge about massage therapy. This knowledge includes, for example, the relationship between massage sites and their therapeutic effects. Massage sites include, for example, acupoints and / or meridians.

[0063] Therefore, if the large-scale language model 310A is given the role of suggesting massages that are appropriate for the user's symptoms (illness) or other aspects of the user's physical condition, then when the large-scale language model 310A acquires data indicating the user's symptoms (illness) or other aspects of the user's physical condition, it can suggest a massage suitable for that user. For example, the large-scale language model 310A can output data indicating acupoints that alleviate the symptoms from data indicating the user's symptoms obtained through a user interview. More specifically, if the user has a headache, the large-scale language model 310A can output data indicating acupoints that alleviate the headache based on its knowledge of the therapeutic effects of acupoints. The large-scale language model 310A can also output what kind of massage technique (kneading, tapping, etc.) should be used to massage those acupoints, or what strength and duration the massage should be performed. In this way, the large-scale language model 310A can have a function (massage suggestion function) that suggests what kind of massage should be performed according to the user's physical condition.

[0064] The large-scale language model 310A of the embodiment utilizes a massage suggestion function and a massage course data generation function that correspond to the user's physical condition to generate massage course data corresponding to the user's physical condition from the input of the user's physical condition. For example, if the user has a headache, the large-scale language model 310A determines an acupoint that will alleviate the headache. The large-scale language model 310A can then generate massage course data that includes a command to move the massage unit 100 to the location of that acupoint and a command to operate the massage unit 100 near that acupoint.

[0065] Furthermore, the large-scale language model 310A of the embodiment can also generate massage course data corresponding to the model of the massage machine 10 based on the model information of the massage machine 10. For example, the massage unit that can massage the location of an acupoint determined by the large-scale language model 310A may differ depending on the model. For example, in one model of massage machine 10, the location of that acupoint may be massaged by a motor-driven massage unit. In this case, the large-scale language model 310A can generate massage course data that includes a command to the motor-driven massage unit to massage the location of that acupoint, based on its knowledge of the location and function of the massage unit of that model of massage machine 10. In another model of massage machine 10, the location of that acupoint may be massaged by a massage unit consisting of air cells. In this case, the large-scale language model 310A can generate massage course data that includes a command to the air cell massage unit to massage the location of that acupoint, based on its knowledge of the location and function of the massage unit of that model of massage machine 10.

[0066] The generation system 300 may include a judge 320. The judge 320 determines whether the massage course data generated by the generator 310 is appropriate. The massage course data generated by the large-scale language model 310A may, in some cases, be inappropriate. Inappropriate massage course data is, for example, one that cannot be executed on the massage machine 10 to which the massage course data is given, or one in which the massage content is inappropriate.

[0067] Massage course data that cannot be executed in the massage machine 10 is, for example, a massage course data that includes commands that the massage machine 10 to which the massage course data is given cannot execute. More specifically, for example, for a massage machine 10 in which the backrest 15 is not equipped with air cells as a massage unit, massage course data that includes commands to operate the air cells of the backrest 15 is not executable.

[0068] Examples of massage course data with inappropriate content include those where the massage course duration is too long. For instance, if the maximum duration of a massage course is set at 15 minutes to avoid the negative effects of prolonged massages, then any massage course data exceeding 15 minutes in duration would be considered to have inappropriate content.

[0069] Another example of inappropriate massage course data is when the massage time on the same body part is too long. If the maximum continuous massage time on the same body part is set to, for example, 3 minutes, then any massage course data where the continuous massage time on the same body part exceeds 3 minutes is considered inappropriate.

[0070] Other inappropriate massage techniques may be set as appropriate in accordance with national regulations regarding massage machines, or rules established by massage machine manufacturers.

[0071] The judgment device 320 may determine whether the generated massage course data is suitable or unsuitable by analyzing the generated massage course data itself, or it may execute the generated massage course data using an emulator that mimics the operation of the massage machine 10, analyze the output of the emulator, and then determine whether it is suitable or unsuitable.

[0072] The judging unit 320 determines whether the massage course data generated by the generator 310 is appropriate. If it is appropriate, the judging unit 320 outputs the massage course data to the generation system 300. If it is not appropriate, the judging unit 320 generates error information. The error information may include, for example, information indicating the unsuitability of the massage course data (e.g., "the duration exceeds 15 minutes"). The error information may also include information indicating how the massage course data should be improved (e.g., "reduce the duration by 1 minute"). In other words, the error information can also be a correction instruction.

[0073] Figure 5 shows the procedure for generating massage course data using the generation system 300 shown in Figure 4.

[0074] In step S501 of Figure 5, the generator 310 acquires input data to be provided to the large-scale language model 310A. The input data is data about the user of the massage machine 10, and is acquired from the massage machine 10, for example, via the network 400.

[0075] The controller 30 of the massage machine 10 functions, for example, as a means of conducting a medical interview with the user using the operating device 50, thereby acquiring information about the user's physical condition. The user's responses during the interview can be made, for example, by operating various operation buttons or by voice input into the microphone 52. In the case of voice input, the ambiguity of human responses (for example, variations in verbal expressions such as "my head hurts," "I have a headache," or "my head is throbbing") may be converted into information in a unified notation or format by AI (generator 310).

[0076] Furthermore, the controller 30 acquires vital data such as the user's heart rate and other user information from the user device 600. The controller 30 may transmit this information as input data to the generator 310. Alternatively, information to be used as input data may be transmitted from the user device 600 to the generator 310. The generator 310 may acquire information to be used as input data from a database containing information about the user's physical condition.

[0077] Furthermore, in step S501, the generator 310 obtains model information of the massage machine 10 that should provide the massage course data. The model information is obtained, for example, from the massage machine 10 that should provide the massage course data. However, if there is only one model, or if the generator 310 already knows the model information, it is not necessary to obtain the model information.

[0078] In step S502, the generator 310 provides the acquired input data (including model information) to the large-scale language model 310A, causing the large-scale language model 310A to output massage course data.

[0079] The determination unit 320 determines whether the massage course data output from the large-scale language model 310A is appropriate (step S503), and if it determines that it is appropriate, it sets the massage course data as the output of the generation system 300 (step S504).

[0080] If the classifier 320 determines that the massage course is inappropriate, it outputs error information (step S505). The error information, along with a regeneration (correction) instruction, is provided to the large-scale language model 310A. Based on the error information (correction instruction), the large-scale language model 310A corrects and regenerates the improved massage course data (step S506). The large-scale language model 310A can repeat the generation process until the classifier 320 determines that the data is appropriate.

[0081] Figure 6 shows another example of the configuration of the generation system 300. Note that, with respect to the generation system 300 in Figure 6, unless otherwise specified, it is the same as the generation system 300 shown in Figure 4.

[0082] The generation system 300 shown in Figure 6 includes an intermediate data generator 311 and a massage course data generator 312 as generators, and generates a massage course in two stages from the input data. The generation system 300 shown in Figure 6 also includes a determination device 320 similar to that in Figure 4.

[0083] The intermediate data generator 311 generates intermediate data from the input data. The generator 311 generates the intermediate data using the first large-scale language model 311A. The first large-scale language model 311A ​​is trained to grasp the user's physical state, etc., from natural language or other inputs that can grasp the user's physical state, etc., and to output intermediate data that indicates the grasped physical state, etc.

[0084] For example, the first large-scale language model 311A ​​is trained with data representing knowledge about massage therapy. Therefore, if the first large-scale language model 311A ​​is given the role of suggesting massages that are appropriate for the user's symptoms (illness) or other aspects of the user's physical condition, then when the first large-scale language model 311A ​​acquires data representing the user's symptoms (illness) or other aspects of the user's physical condition, it can output a massage that is suitable for that user.

[0085] For example, if a user has a headache, the first large-scale language model 311A ​​can output intermediate data that includes data indicating acupoints that can alleviate the headache, based on its knowledge of massage therapy. The first large-scale language model 311A ​​can also output information on what massage technique (kneading, tapping, etc.) should be used to massage those acupoints, or what intensity and duration of massage should be applied. In this way, the first large-scale language model 311A ​​can output intermediate data that suggests what kind of massage should be performed according to the user's physical condition.

[0086] The massage course data generator 312 generates massage course data from intermediate data. The generator 312 generates massage course data using the second large-scale language model 312A. The second large-scale language model 312A is trained with data representing knowledge of the programming language executed by the processor 30A of the controller 30 of the massage machine 10. Furthermore, the second large-scale language model 312A is trained with data representing knowledge for operating the massage unit and other functions controllable by the controller 30 in the massage machine 10, so that the operation of the massage machine 10, such as massage movements, can be expressed by program. In addition, the second large-scale language model 312A is trained with a large amount of massage course data. Moreover, the second large-scale language model 312A is trained with data representing knowledge of the functions of various models of massage machine 10.

[0087] Therefore, when the second large-scale language model 312A is given input indicating what kind of massage should be performed (for example, which body parts (acupoints, etc.) should be massaged), it can generate massage course data to operate the massage machine 10 in accordance with that input. Furthermore, when the second large-scale language model 312A is given model information, it can also generate massage course data corresponding to the model of the massage machine 10 based on the model information of the massage machine 10.

[0088] Thus, the second large-scale language model 312A is trained to generate massage course data for each machine model. Here, the intermediate data, which is the output of the first large-scale language model 311A, has relatively low dependency on the massage machine model and is highly generalizable. On the other hand, the massage course data, which is the output of the second large-scale language model 312A, is machine-dependent. Therefore, it is preferable to use the first large-scale language model 311A ​​and the second large-scale language model 312A separately, as this allows for efficient training of each model.

[0089] Figure 7 shows the procedure for generating massage course data using the generation system 300 shown in Figure 6.

[0090] In step S701 of Figure 7, the generator 311 acquires input data to be provided to the first large-scale language model 311A. The input data is data about the user of the massage machine 10, and is acquired from the massage machine 10, for example, via the network 400.

[0091] In step S702, the generator 311 provides the acquired input data (including model information) to the first large-scale language model 311A, causing the first large-scale language model 311A ​​to output intermediate data. The intermediate data includes, for example, data indicating acupoints that should be massaged according to the user's physical condition (symptoms) indicated by the input data.

[0092] In step S703, the generation system 300 presents intermediate data to the user, accepts additional data from the user as needed (such as data indicating other areas to be massaged), adds the additional data to the intermediate data, or modifies the intermediate data based on the additional data.

[0093] In step S704, the generator 312 acquires intermediate data and provides it as input to the second large-scale language model 312A. The second large-scale language model 312A outputs massage course data corresponding to the intermediate data. For example, the second large-scale language model 312A generates massage course data that includes massage actions for the locations of acupoints included in the intermediate data.

[0094] The determination unit 320 determines whether the massage course data output from the second large-scale language model 312A is appropriate (step S705), and if it determines that it is appropriate, it sets the massage course data as the output of the generation system 300 (step S706).

[0095] If the classifier 320 determines that the massage course is inappropriate, it outputs error information (step S707). The error information, along with a regeneration (correction) instruction, is provided to the second large-scale language model 312A. Based on the error information (correction instruction), the second large-scale language model 312A corrects and regenerates the improved massage course data (step S708). The second large-scale language model 312A can repeat the generation process until the classifier 320 determines that the data is appropriate. Repeating the generation process in the second large-scale language model 312A in Figure 6 is preferable because it requires less computation than repeating the generation process in the large-scale language model 310A in Figure 4.

[0096] Figures 8 and 9 show an example of a massage sequence according to a generated massage course. The sequence shown in Figure 8 consists of multiple actions, such as the first action S1, the second action S2, the third action S3, ..., and the nth action SN (where N is any integer), arranged in a predetermined order.

[0097] The first operation S1 is, for example, a shoulder position detection operation. The shoulder position detection operation (step S1) may include, for example, a substep S1-1 in which the first massage unit 100 rises from an initial position toward shoulder position 1A (see Figure 3). The initial position in step S1-1 is, for example, the lowest position in direction Y. In step S1-1, as the massager 150 rises while being applied to the back of the person being treated 1, the load that the massager 150 receives from the person being treated 1 decreases when the massager 150 reaches shoulder position 1A. The controller 30 can detect the position of the massage unit 100 when the decrease in load is detected by the sensor as shoulder position 1A. Based on the detected shoulder position 1A, the controller 30 can determine different body types and massage positions (for example, the positions of acupuncture points included in the massage course) for each person being treated 1.

[0098] As shown in Figure 9, the shoulder position detection operation (step S1) accepts a position adjustment operation for the detected shoulder position 1A (see Figure 3), and may include step S1-2 in which a kneading operation is performed at the detected shoulder position with a "medium" intensity. The patient 1 can perform an operation to fine-tune the detected shoulder position 1A during the kneading operation in step S1-2.

[0099] The second action S2 is, for example, a tapping motion in a range A1 that includes acupoint B1, as determined by a large-scale language model. As shown in Figure 8, the range A1 that includes acupoint B1 is a range of the body near the body position where acupoint B1 is located. This tapping motion (step S2) may consist of multiple substeps, such as substep S2-(M-1) which moves back and forth across range A1 while tapping with "medium" force, or substep S2-M which moves to range A2 that includes acupoint B2 while tapping with "medium" force, as shown in Figure 9.

[0100] The third action S3 is, for example, a kneading motion on a range A2 that includes acupoint B2, which has been determined by a large-scale language model. As shown in Figure 8, the range A2 that includes acupoint B2 is a range of the body near the body position where acupoint B1 is located. This kneading motion (step S3) may consist of steps S3-1 and S3-3, in which the body moves back and forth across range A2 while kneading with a "medium" intensity, and steps S3-2 and S3-4, in which the body moves back and forth across range A2 while kneading with a "weak" intensity. In other words, the third action S3 involves kneading range A2 while moving back and forth up and down multiple times, with the kneading intensity changing with each movement.

[0101] The first operation S1, second operation S2, third operation S3, etc., set in the sequence may also include the operation of the air cell 201.

[0102] [Other embodiments] In the above explanation, the large-scale language model 310A outputs what kind of massage should be given to the user (for example, which body parts (acupoints, etc.) should be massaged) based on the input data provided, and generates massage course data. However, it is also possible to output what kind of massage should not be given to the user (for example, which body parts (acupoints, etc.) should not be massaged) based on the input data, and generate massage course data.

[0103] For example, the input data provided to the large-scale language model 310A includes age-related data regarding the user's age. When obtaining information about the user's physical condition through a medical interview conducted by the operating device 50, the input data can include age-related data by including the user's age in the content of the interview (step S501 in Figure 5). As a result, input data including age-related data is transmitted to the generator 310. The generator 310 provides the acquired input data (including age-related data) to the large-scale language model 310A, causing the large-scale language model 310A to output massage course data (step S502 in Figure 5). In this way, it is possible to generate massage course data that avoids massaging parts of the user that should not be massaged according to the user's age. Alternatively, the input may be provided from the user device 600 instead of the operating device 50. The following description will focus on the case where generator 310 is used, but the same principles apply when intermediate data generator 311 and massage course data generator 312 are used instead of generator 310. In other words, the intermediate data may be configured to include data indicating which parts of the user should not be massaged, depending on the input data, which includes age-related data.

[0104] Furthermore, the age-related data that users enter in the questionnaire does not have to be their exact age. For example, it could be an age group. Age groups include categories such as elderly, adults, and children, or age groups such as those in their 20s and 30s. This approach can alleviate users' reluctance to enter their own age.

[0105] Furthermore, age-related data can be inferred from the content of the medical questionnaire. For example, if the questionnaire reveals that the user has chronic lower back pain or weak bones, the system can infer that the user is elderly. In this way, age-related data can be obtained indirectly without requiring the user to input their age.

[0106] One area that should not be massaged is the lower back, especially in elderly individuals. If the lower back is curved due to aging or if the person suffers from osteoporosis, massaging the lower back can potentially injure the muscles or even cause a fracture, so lower back massage should be avoided. Therefore, if the large-scale language model 310A includes information that the user is elderly, it generates massage course data that avoids massaging the lower back. This massage course data that avoids massaging the lower back may simply exclude the lower back massage, or it may involve massaging other body parts instead.

[0107] Furthermore, adjustment data may be generated to adjust massage movements not only for areas that should not be massaged, but also according to age-related data included in the input data, and massage course data may be generated based on the adjustment data. For example, the adjustment data adjusts the strength of the massage movement according to the age-related data included in the input data. If the input data includes elderly people as age-related data, the strength of the massage movement is adjusted to be weaker overall. Specifically, the large-scale language model 310A generates adjustment data that causes the controller 30 to control the maximum strength of the massage movement when the input data includes elderly people to be limited to a strength equivalent to "medium" in the normal massage movement strength (for example, the strength of the massage movement when the user is an adult). In this way, even if an elderly user accidentally tries to massage with excessive strength, the maximum strength of the massage movement is controlled to be limited to a strength equivalent to "medium" for adults, thus reducing the possibility of muscle injury or fracture due to excessive stimulation. Note that the above strength is just an example and is not limited to it.

[0108] Similarly, if the input data includes age-related data such as children, adjustment data may be generated to control the controller 30 to generally reduce the intensity of the massage motion. This prevents excessive stimulation to children whose muscles and bones are still immature.

[0109] Furthermore, the adjustment data may be modified to change the massage technique depending on the age-related data included in the input data. For example, if the large-scale language model 310A includes elderly people as age-related data in the input data, it may generate adjustment data that uses rubbing instead of kneading or tapping as the massage technique. In this way, it is possible to prevent bones that have weakened with age from fracturing due to kneading or tapping.

[0110] Similarly, if the input data includes age-related data such as children, the system may generate adjustment data to perform rubbing instead of kneading and tapping as the massage technique. This prevents excessive stimulation to children whose muscles and bones are still immature.

[0111] Furthermore, the adjustment data may be modified to change the massage time to the same area during a massage action, depending on the age-related data included in the input data. For example, if the large-scale language model 310A includes elderly individuals as age-related data in the input data, it generates adjustment data that shortens the normal massage time to the same area during a massage action (e.g., 3 minutes) to a shorter time (e.g., 2 minutes). In this way, the risk of muscle injury or fracture in elderly individuals with weakened muscles and bones due to continuous stimulation of the same area can be reduced.

[0112] Similarly, if the input data includes age-related data such as children, adjustment data may be generated to shorten the massage time for the same area during the massage motion. This prevents excessive stimulation to children whose muscles and bones are still immature.

[0113] Furthermore, the adjustment data may be modified to change the duration of the massage course depending on the age-related data included in the input data. For example, the large-scale language model 310A generates adjustment data that shortens the duration of the massage course from the normal time (e.g., 15 minutes) to a shorter time (e.g., 10 minutes) if the age includes elderly people. This prevents excessive strain on elderly people with reduced physical strength.

[0114] Similarly, if the input data includes age-related data such as children, adjustment data may be generated to shorten the duration of the massage course. This allows the massage course to be completed before the user experiences stress from prolonged stimulation or maintaining the same posture.

[0115] While age-related data was used as an example above, the system is not limited to this. Massage course data may be generated based on adjustment data that modifies areas that should not be massaged and / or massage movements, using the user's symptoms (medical condition) or other aspects of the user's physical condition.

[0116] Furthermore, users may be grouped by combining multiple input data such as age, weight, and height. For example, users could be divided into a group with strong (or normal) bones and a group with weak bones. If the age-related data includes elderly individuals, they would be classified into the weak bones group. Even if the age-related data does not include elderly individuals, if a user's weight is too low relative to their height, they would be classified into the weak bones group. In this way, users may be grouped, and massage course data may be generated based on adjustment data that adjusts the areas that should not be massaged and / or the massage movements accordingly.

[0117] The present invention is not limited to the above embodiments, and various modifications are possible. [Explanation of Symbols]

[0118] 1: Patient receiving treatment 1A: Shoulder position 10: Massage machine 11: Main unit of the device 13: Seat part 15: Backrest 17: Footrest 21: Reclining Actuator 23: Oscillating Actuator 30: Controller 30A: Processor 30B: Storage device 30C: Communication Department 30D: Computer Program 31: Massage course data 32: Massage course data 40: Sensor 50: Operating device 51: Display 52: Mike 100: First Massage Department 120: Base Unit 121: Motor with adjustable power level 130: Advance / Retreat Unit 131: Massage motor 132: Striking motor 140: Arm 150: Treatment child 160: Movement mechanism 161: Lifting motor 162: Linear Actuator 200: Second Massage Department 201: Air Cell 203: Air Circuit 300: Generation System 300A: Processor 300B: Storage device 300C:Communication Department 310: Massage course data generator 310A: Large-scale language models 311: Intermediate data generator 311A: First Large-Scale Language Model 312: Massage course data generator 312A: Second Large-Scale Language Model 320: Judgment device 400: Network 500: Massage System 600: User device

Claims

1. Input data regarding the user of the massage machine is input into a large-scale language model that generates massage motion data for causing the massage machine to perform massage operations. Massage motion data corresponding to the input data is output from the large-scale language model. Based on the massage motion data output from the large-scale language model, the massage machine is made to perform a massage motion. Massage system.

2. The large-scale language model is configured to generate massage operation data corresponding to the functions of the massage section of the massage machine, based on the model information of the massage machine. The massage system according to claim 1.

3. The aforementioned large-scale language model, The first large-scale language model and, The second large-scale language model, Equipped with, The first large-scale language model is configured to generate intermediate data from the input data, which is then given as input to the second large-scale language model. The second large-scale language model is configured to generate the massage motion data from the intermediate data. The massage system according to claim 1.

4. The second large-scale language model is configured to generate massage operation data corresponding to the functions of the massage section of the massage machine, based on the model information of the massage machine. The massage system according to claim 3.

5. The aforementioned intermediate data includes area data indicating the parts of the user that should be massaged. The first large-scale language model generates the intermediate data, including the region data, from the input data. The second large-scale language model generates the massage motion data from the intermediate data, which includes the body part data. The massage system according to claim 3.

6. The second large-scale language model is configured to generate massage operation data that causes the massage unit of the massage machine corresponding to the model information to perform a massage operation on the body part indicated by the body part data, based on the model information and body part data of the massage machine. The massage system according to claim 5.

7. The appropriateness of the massage motion data output from the large-scale language model is determined. The massage system according to claim 1.

8. The system determines whether the massage motion data output from the large-scale language model is appropriate, and if it is determined to be inappropriate, it causes the large-scale language model to output corrected massage motion data. The massage system according to claim 1.

9. The aforementioned input data includes age-related data concerning the user's age. The massage motion data output from the large-scale language model is configured to generate at least one of the following based on the age-related data: body part data indicating body parts of the user that should not be massaged, and adjustment data for adjusting the massage motion. The massage system according to claim 1.

Images