Exercise support device, exercise support system, exercise support method, and exercise support program

Abstract

To provide an exercise support device that allows an exercise to be done at an optimum exercise strength in consideration of individual differences (less influenced by individual differences), and allows the exercise to be done in a simple facility at less cost and without requiring an expert's presence.SOLUTION: An exercise support device for supporting a user doing an exercise at an optimum exercise strength includes: an acquisition part for acquiring bioelectric impedance of the user doing an exercise; a conversion part for subjecting the acquired bioelectric impedance to Fourier transformation and generating a frequency spectrum; a calculation part for calculating a difference between current exercise strength and the optimum exercise strength of the user doing exercise when the generated frequency spectrum is input; and an output part for outputting a difference between the current exercise strength and the calculated optimum exercise strength. The calculation part includes a learned calculation model subject to machine learning using teacher data so as to calculate a difference between the current exercise strength and the optimum exercise strength of the user doing exercise when the frequency spectrum is input.SELECTED DRAWING: Figure 11

Description

[Technical Field]

[0001] The present disclosure relates to an exercise support device, an exercise support system, an exercise support method, and an exercise support program that support a user in exercising at an optimal exercise intensity. [Background technology]

[0002] Moderate exercise is effective for maintaining good health. Exercise that is too light will not provide sufficient benefits, while excessive exercise can be harmful to health. However, it is not easy to accurately determine the appropriate exercise intensity.

[0003] In rehabilitation settings, the optimum exercise intensity is considered to be near the anaerobic threshold (AT) determined by a cardiopulmonary exercise test (CPX) using expiratory gas analysis (Patent Document 1).

[0004] As simple indicators of exercise intensity, the Karvonen method, which calculates based on age and resting heart rate, and the Borg index (Patent Document 2), which is based on subjective symptoms, have been proposed. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] Japanese Patent Application Publication No. 2020-130640 [Patent Document 2] Japanese Patent Application Laid-Open No. 2015-186542 Summary of the Invention [Problem to be solved by the invention]

[0006] However, cardiopulmonary exercise testing (CPX) using expiratory gas analysis requires expensive specialized equipment and the presence of specialists such as medical staff who can determine AT values, which places a heavy burden on users, including the high cost of the test.

[0007] The Karvonen method and the Borg index, which is based on subjective symptoms, have problems such as not taking individual differences into sufficient consideration and not being an objective, quantitative indicator.

[0008] Therefore, there is a demand for an exercise support device that can be implemented with simple equipment, does not require the presence of an expert to determine AT, thereby reducing the cost burden, and enables exercise at the optimal exercise intensity that takes individual differences into account. [Means for solving the problem]

[0009] The gist of the present invention is as follows. (1) An exercise support device that supports a user in exercising at an optimal exercise intensity, an acquisition unit that acquires bioelectrical impedance of the user during exercise; a transform unit that performs a Fourier transform on the acquired bioelectrical impedance to generate a frequency spectrum; a calculation unit that receives the generated frequency spectrum and calculates a difference between the current exercise intensity of the user and an optimal exercise intensity of the user; an output unit that outputs the difference between the calculated optimal exercise intensity and the current exercise intensity; Equipped with The calculation unit has a trained calculation model that has been subjected to machine learning using training data so as to calculate a difference between the current exercise intensity of the user and an optimal exercise intensity when the frequency spectrum is input. Exercise support device. (2) The exercise support device according to (1) above, wherein the difference between the optimal exercise intensity and the current exercise intensity is the difference between the heart rate at the AT of the user during exercise and the current heart rate of the user during exercise. (3) The exercise support device according to (1) or (2) above, further comprising a display unit that displays the difference between the current exercise intensity and the optimal exercise intensity output from the output unit. (4) The trained calculation model is The input data is a frequency spectrum generated by Fourier transforming the bioelectrical impedance measured for the subject during exercise for each time series data; and The difference in exercise intensity of the subject during the exercise compared to the exercise intensity at AT measured by a cardiopulmonary exercise stress test using expiratory gas analysis, which is performed simultaneously with the measurement of bioelectrical impedance of the subject, is used as an output label. The exercise support device according to any one of (1) to (3) above, which has been subjected to learning by the above method. (5) An exercise support device according to any one of (1) to (4) above; an impedance measuring device for measuring the bioelectrical impedance of the user during exercise; An exercise support system comprising: (6) The exercise support system according to (5) above, wherein the measurement device includes a four-terminal probe measurement unit. (7) The exercise support system according to (6) above, wherein the four-terminal probe measurement unit includes two electrodes arranged on each of the left and right handlebars of the fitness bike. (8) a receiving unit that receives the difference between the current exercise intensity and the optimal exercise intensity transmitted from the output unit; and a display unit that displays whether the current exercise intensity of the user during exercise is higher, the same as, or lower than the optimal exercise intensity of the user, based on the difference between the current exercise intensity and the optimal exercise intensity received by the receiving unit. The exercise support system according to any one of (5) to (7) above, comprising: (9) An exercise support method for supporting a user to exercise at an optimal exercise intensity, acquiring bioelectrical impedance of the user during exercise; Fourier transforming the acquired bioelectrical impedance to generate a frequency spectrum; and inputting the generated frequency spectrum to calculate a difference between the user's current exercise intensity and an optimal exercise intensity during the exercise; outputting a difference between the calculated optimal exercise intensity and the current exercise intensity; Including, the calculating uses a trained calculation model that has been subjected to machine learning using training data so as to calculate a difference between the current exercise intensity of the user and an optimal exercise intensity when the frequency spectrum is input; Exercise support method. (10) An exercise support program that supports a user in exercising at an optimal exercise intensity, a transform function that performs a Fourier transform of the bioelectrical impedance measured during the user's exercise to generate a frequency spectrum; and a calculation function for calculating a difference between the current exercise intensity and the optimal exercise intensity of the user when the generated frequency spectrum is input; Including, the calculation function has a trained calculation model that has been subjected to machine learning using training data so as to calculate a difference between the optimal exercise intensity of the user during exercise and the current exercise intensity when the frequency spectrum is input; Exercise support program. [Effects of the Invention]

[0010] According to the present disclosure, it is possible to provide an exercise support device that can be implemented with simple equipment, does not require the presence of an expert, is inexpensive, and enables exercise at optimal exercise intensity taking into account individual differences. [Brief explanation of the drawings]

[0011] [Figure 1] FIG. 1 is a schematic diagram showing an example of the configuration of an exercise support system 100 including the exercise support device 10. As shown in FIG. [Figure 2] FIG. 2 is a diagram showing an example of the configuration of the exercise support device 10. As shown in FIG. [Figure 3] FIG. 3 shows an example of a frequency spectrum obtained by Fourier transforming the bioelectrical impedance measured on a subject. [Figure 4] FIG. 4 is a configuration diagram of an example of a training data collection system for collecting training data used in machine learning. [Figure 5]FIG. 5 is a configuration diagram of an example of a machine learning system when machine learning is performed. [Figure 6] FIG. 6 is a schematic diagram showing an example of a machine learning algorithm. [Figure 7] FIG. 7 is a schematic diagram showing an example of the calculation model M1. [Figure 8] FIG. 8 is a schematic diagram showing another example of the calculation model M1. [Figure 9] FIG. 9 is a diagram illustrating an example of a configuration of a support system when a user uses the exercise support device 10. As shown in FIG. [Figure 10] FIG. 10 is a diagram showing an example of the operational flow of the calculation process by the conversion processing unit 152, the calculation processing unit 154, and the output processing unit 155. [Figure 11] FIG. 11 shows a measurement example in which the exercise support device 10 estimates the difference between the AT heart rate and the current heart rate for 15 users using a neural network as a trained calculation model. [Figure 12] FIG. 12 shows the results of 10-fold cross-validation of data estimated by the exercise support device 10 using a neural network as a calculation model. [Figure 13] FIG. 13 shows the results of 10-fold cross-validation of data estimated by the exercise support device 10 using a random forest as a learning model. [Figure 14] FIG. 14 is a schematic diagram showing the appearance of a measurement unit in which electrode pads having a four-terminal configuration are arranged on the surface of a substantially round bar. [Figure 15] Figure 15 shows a graph of the current HR (heart rate) estimates for 15 users using a trained calculation model that generated an output label as HR (heart rate) measured by CPX. [Figure 16] Figure 16 shows a graph of the current VO2 (oxygen uptake) estimated for 15 users using a trained calculation model that generated output labels as VO2 (oxygen uptake) measured by CPX. [Figure 17]Figure 17 shows a graph of the current VE / VO2 (pulmonary ventilation / oxygen uptake) estimated for 15 users using a trained calculation model that generated output labels using the VE / VO2 (pulmonary ventilation / oxygen uptake) measured by CPX. [Figure 18] Figure 18 shows a graph of the current VE / VO2 (pulmonary ventilation / oxygen uptake) estimated for 15 users using a trained calculation model that generated output labels using VE / VO2 (pulmonary ventilation / oxygen uptake) measured by CPX. [Figure 19] Figure 19 shows a graph of the current R (gas exchange ratio) estimated for 15 users using a trained calculation model that generated output labels as R (gas exchange ratio) measured by CPX. [Figure 20] Figure 20 shows a graph of the current AT (0, 1 decision) estimated for 15 users using a trained calculation model that generated output labels as AT (0, 1 decision) measured by CPX. [Figure 21] FIG. 21 is a schematic cross-sectional view of a measurement unit in which electrode pads having a four-terminal configuration are arranged on the surface of a substantially round bar. DETAILED DESCRIPTION OF THE INVENTION

[0012] Various embodiments of the present invention will be described below with reference to the drawings. However, it should be noted that the technical scope of the present invention is not limited to these embodiments, but extends to the inventions set forth in the claims and their equivalents.

[0013] The present disclosure relates to an exercise support device that supports a user in exercising at an optimal exercise intensity, comprising: an acquisition unit that acquires the bioelectrical impedance of the user while exercising; a conversion unit that performs a Fourier transform on the acquired bioelectrical impedance to generate a frequency spectrum; a calculation unit that, when the generated frequency spectrum is input, calculates a difference between the user's current exercise intensity and the optimal exercise intensity while exercising; and an output unit that outputs the difference between the current exercise intensity and the calculated optimal exercise intensity, wherein the calculation unit has a learned calculation model that has been subjected to machine learning using training data so as to calculate the difference between the user's current exercise intensity and the optimal exercise intensity while exercising when the frequency spectrum is input.

[0014] This device makes it easier than ever to determine an individual's anaerobic threshold (hereinafter referred to as AT) and exercise accordingly. More specifically, this device allows users to exercise at an optimal exercise intensity that takes individual differences into account, with simple equipment and no need for expert supervision, reducing costs. In particular, this device can measure the relative relationship between the user's optimal exercise intensity and the current exercise intensity during exercise in real time, enabling continuous exercise at the user's optimal exercise intensity.

[0015] In the present application, the optimal exercise intensity refers to an exercise intensity near the AT measured by a cardiopulmonary exercise test using expiratory gas analysis (hereinafter also referred to as CPX). The exercise intensity near the AT is, for example, substantially the AT itself, or a predetermined range around the AT, such as AT ±1%, AT ±3%, AT ±5%, or AT ±10%. The exercise intensity near the AT may also be a predetermined range below the AT or below the AT, or a predetermined range above the AT or above the AT. The predetermined range below the AT or below the AT is, for example, within the range of below the AT or below the AT and AT −5%, AT −10%, AT −15%, or AT −20%. The predetermined range above the AT or above the AT is, for example, within the range of above the AT or above the AT and AT +5%, AT +10%, AT +15%, or AT +20%.

[0016] In CPX, a subject exercises on an exercise machine such as a fitness bike while gradually increasing the exercise load (pedaling, etc.), and their breathing volume, oxygen intake (VO2), and carbon dioxide excretion (VCO2) are measured using a breath gas analyzer. Based on the measurement results, an expert can determine the subject's AT. Examples of AT indicators include heart rate (HR), pulse rate, VO2 (oxygen intake), VE / VO2 (pulmonary ventilation / oxygen intake), VE / VCO2 (pulmonary ventilation / CO2 output), R (gas exchange ratio), or the load of the exercise machine. Preferably, heart rate (HR) is used. The load of the exercise machine is, for example, the pedal load of the exercise bike. The exercise bike can be, for example, an Aero Bike (registered trademark).

[0017] 1 is a schematic diagram showing an example of the configuration of an exercise support system 100 including the exercise support device 10. The exercise support device 10 calculates the difference between the user's current exercise intensity and the optimal exercise intensity during exercise based on the bioelectrical impedance measured for the user during exercise. The current exercise intensity is the exercise intensity of the user during exercise in real time.

[0018] The exercise support device 10 can be connected to an impedance measuring device that measures the bioelectrical impedance of a user or subject during exercise. The impedance measuring device includes an impedance meter 20 connected to the exercise support device 10 and a measuring unit 30 connected to the impedance meter 20 and in contact with the user or subject. The user is a person who intends to exercise at an optimal exercise intensity using the exercise support device 10, and the subject is a person who obtains training data from the exercise support device 10. The exercise support device 10, the impedance meter 20, and the measuring unit 30 can be connected via a wired or wireless connection that allows for the transmission and reception of measurement data.

[0019] The exercise support device 10 calculates the difference between the user's current exercise intensity and the optimal exercise intensity based on the user's bioelectrical impedance measured by an impedance measuring device including a measuring unit 30 and an impedance meter 20, and outputs the calculated difference between the current exercise intensity and the optimal exercise intensity. The user can increase, decrease, or maintain the current exercise intensity based on the difference between the current exercise intensity and the optimal exercise intensity. More specifically, the user can visually recognize the difference between the displayed optimal exercise intensity and the current exercise intensity and adjust their own exercise intensity or the load of the exercise equipment, or the exercise equipment can automatically adjust the load so that the current exercise intensity approaches the optimal exercise intensity. In this way, the exercise support device 10 allows the user to continuously exercise at an exercise intensity close to the optimal exercise intensity.

[0020] (Impedance measuring device) A measuring unit 30 connected to an impedance meter 20 is worn by or comes into contact with the user or subject. The measuring unit 30 may be attached to the user or subject, or may be provided on exercise equipment such as a fitness bike. The bioelectrical impedance of the user or subject during exercise measured by the measuring unit 30 and the impedance meter 20 can be acquired by an acquisition unit of the exercise support device 10.

[0021] The measuring unit 30 includes electrodes, and as long as it can measure the bioelectrical impedance of a user or subject during exercise, the manner in which the electrodes are attached to or contacted with the user is not limited. For example, the electrodes may be attached to the user or subject, or the user or subject may contact the electrodes by gripping or touching them. From the perspective of simple and accurate measurement, the electrodes are preferably attached to the handles of exercise equipment and contacted with the user or subject by the user or subject gripping them. The parts of the user's body that come into contact with the measuring unit 30 are preferably two or more locations through which the current passes through the user's chest, and more preferably both palms, where the skin is thin and contact can be easily maintained by gripping. By contacting the measuring unit 30 with two or more locations through the chest, the movement of the lungs and heart can be detected with greater sensitivity, enabling more accurate measurement of bioelectrical impedance.

[0022] The measurement unit 30 can be equipped with conventional resistance measurement functions such as the two-terminal method, four-terminal method, and five-terminal method. The measurement unit 30 preferably includes a four-terminal probe. A four-terminal method measurement unit 30 equipped with a four-terminal probe includes a current source terminal for supplying a constant current and a voltage detection terminal for detecting a voltage drop. Because the input impedance of the voltmeter is high and almost no current flows through the lead wire on the voltage detection terminal side, the four-terminal method can accurately measure the bioelectrical impedance of a user or subject without being affected by the conductor resistance or contact resistance of the measurement leads. The constant current is preferably an AC current with an amplitude of 100 μA or less and a frequency of 0.1 Hz or more, or a DC current of 10 μA or less, and more preferably an AC current with an amplitude of 100 μA or less and a frequency of 0.1 Hz or more. Using the above-mentioned preferred constant current enables more accurate measurement of bioelectrical impedance.

[0023] The measuring unit 30 preferably has a generally round rod shape with electrode pads arranged on its surface. The measuring unit, which consists of two roughly round rods with electrode pads arranged on their surfaces, can be held by the user or subject in each hand, allowing for easy measurement of the user's or subject's bioelectrical impedance during exercise.

[0024] More preferably, the measuring unit 30 has a generally round rod shape with electrode pads configured for the four-terminal method arranged on its surface. FIG. 14 shows a schematic view of the appearance of a measuring unit in which electrode pads configured for the four-terminal method are arranged on the surface of a generally round rod. FIG. 21 is a schematic cross-sectional view of the measuring unit 30 in which electrode pads 31 and 32 configured for the four-terminal method are arranged on the surface of a generally round rod 33. The two electrode pads are arranged on the surface of the generally round rod so that when the measuring unit is held in each hand, the two electrode pads 31 and 32, which are spaced apart from each other, come into contact with the skin of the hands. The electrode pad 31 is a current application electrode (current source terminal) that supplies a constant current, and the electrode pad 32 is a voltage measurement electrode (voltage detection terminal) that detects a voltage drop.

[0025] The electrode pad is preferably a metal-plated nonwoven fabric. Because the metal-plated nonwoven fabric is conductive and flexible, it can be easily placed on the surface of a generally round rod of any shape. It can also absorb sweat produced by an exercising user or subject, thereby reducing discomfort to the user or subject. The plating can be, for example, copper, nickel, or a combination thereof. More preferably, the electrode pad is a metal-plated nonwoven fabric with adhesive properties. The adhesive properties of the electrode pad allow it to be easily placed on the surface of a generally round rod or the like. For example, the TR-35NH manufactured by Takeuchi Industries Co., Ltd. can be used as the electrode pad.

[0026] Measurement data of the bioelectrical impedance of a user or subject during exercise is likely to contain noise, but by providing the measurement unit 30 with a four-terminal measurement circuit that is shaped to be easy for the user or subject to hold, bioelectrical impedance can be measured accurately and easily.

[0027] The measuring unit 30 is preferably provided on the handlebars of the exercise bike. The measuring unit 30 has a four-terminal probe with four electrodes, two on each handlebar of the exercise bike, and the user or subject holds the measuring unit 30 while pedaling the exercise bike, allowing accurate bioelectrical impedance measurements to be taken even while the user or subject is exercising. The impedance meter 20 can be any conventional, commonly used instrument.

[0028] (Exercise support device) The exercise support device 10 is an information processing device such as a computer, a server, etc. The exercise support device 10 has a calculation function for calculating the current exercise intensity relative to the optimal exercise intensity of the user based on the bioelectrical impedance of the user during exercise acquired by an acquisition unit.

[0029] The exercise support device 10 may be configured as a single information processing device, or may be a collection of multiple physically separate information processing devices. In this case, each of the multiple information processing devices may have the same function, or may have the functions of a single exercise support device 10 in a distributed manner.

[0030] FIG. 2 is a diagram showing an example of the configuration of exercise support device 10. Exercise support device 10 has a receiving function for acquiring data on the bioelectrical impedance of a user during exercise measured by an impedance measuring device, and a converting function for converting the bioelectrical impedance data into a frequency spectrum. Exercise support device 10 also has a learning function for training a calculation model, a calculation function for calculating the difference between the user's current exercise intensity and the optimal exercise intensity during exercise using the learned calculation model, and an output function for outputting the results of calculation by the calculation function. To this end, exercise support device 10 includes a communication unit 11, a memory unit 12, a display unit 13, an operation unit 14, and a processing unit 15.

[0031] The communication unit 11 is implemented as hardware, firmware, communication software such as a TCP / IP driver or a PPP driver, or a combination thereof. The communication unit 11 can be implemented via wireless or wired communication. The exercise support device 10 can acquire bioelectrical impedance data from the impedance measuring device via the communication unit 11. The communication unit 11 may receive bioelectrical impedance data from the impedance measuring device via serial communication using a USB cable. The communication unit 11 may also have an interface circuit for short-range wireless communication according to a communication method such as Bluetooth (registered trademark) and may receive radio waves from the impedance measuring device. The communication unit 11 may also have a receiving circuit for receiving various signals corresponding to the bioelectrical impedance data via infrared communication or the like. The communication unit 11 may also have a communication interface circuit for a wired LAN. This allows the exercise support device 10 to acquire bioelectrical impedance from the impedance measuring device via the communication unit 11.

[0032] Exercise support device 10 may include an input / output device that detachably holds a portable storage medium instead of or together with communication unit 11. In this case, the input / output device acquires bioelectrical impedance data stored in the portable storage medium and supplies the acquired bioelectrical impedance data to processing unit 15. This allows exercise support device 10 to acquire bioelectrical impedance from the impedance measuring device via the portable storage medium.

[0033] The storage unit 12 is, for example, a semiconductor memory device such as a ROM or RAM. The storage unit 12 may also be, for example, a magnetic disk, an optical disk, or any other storage device capable of storing data. The storage unit 12 stores an operating system program, a driver program, an application program, data, and the like used for processing in the processing unit 15. The computer programs stored in the storage unit 12 may be installed into the storage unit 12 from a computer-readable portable recording medium such as a CD-ROM or a DVD-ROM using a known setup program or the like.

[0034] The data stored in the storage unit 12 includes a calculation model M1, training data TD, etc., which will be described later. The storage unit 12 also stores bioelectrical impedance data acquired from an impedance measuring device. The storage unit 12 may also temporarily store data related to predetermined processing.

[0035] The display unit 13 is a display. The display unit 13 may be a liquid crystal display, an organic EL display, etc. The display unit 13 displays a frequency spectrum based on the bioelectrical impedance data supplied from the processing unit 15, information on the results of calculation of the difference between the user's optimal exercise intensity and the current exercise intensity, etc.

[0036] The operation unit 14 can be a keyboard, a mouse, and / or a pointing device such as a touch panel. A user can use the operation unit 14 to input letters, numbers, symbols, or positions on the display screen of the display unit 13. When operated by a user, the operation unit 14 generates a signal corresponding to the operation. The generated signal is then supplied to the processing unit 15 as an instruction from the user.

[0037] The processing unit 15 is a processing device that loads the operating system program, driver program, and application program stored in the storage unit 12 into memory and executes instructions contained in the loaded programs. The processing unit 15 is, for example, an electronic circuit such as a CPU, an MPU, or a DSP, or a combination of various electronic circuits. Although the processing unit 15 is illustrated in FIG. 2 as a single component, the processing unit 15 may be a collection of multiple physically separate processors.

[0038] The processing unit 15 executes various commands included in an application program (control program) stored in the storage unit 12, thereby functioning as a reception processing unit 151, a conversion processing unit 152, a learning processing unit 153, a calculation processing unit 154, and an output processing unit 155. Below, examples of the functions of the reception processing unit 151, the conversion processing unit 152, the learning processing unit 153, the calculation processing unit 154, and the output processing unit 155 will be described with reference to FIGS. 3 to 9.

[0039] The receiving processor 151 receives the bioelectrical impedance data transmitted from the impedance measuring device via the communication unit 11 and stores the received bioelectrical impedance data in the storage unit 12.

[0040] The transformation processing unit 152 performs a Fourier transform on the bioelectrical impedance data stored in the storage unit 12 to generate a frequency spectrum, and stores the frequency spectrum in the storage unit 12. The Fourier transform of the bioelectrical impedance data can be performed for each time series data.

[0041] The Fourier transform for each time series data includes Fourier transforming data for each second predetermined time while shifting the data by a first predetermined time to generate a frequency spectrum. The first predetermined time can be, for example, 1 to 10 seconds. The second predetermined time can be, for example, 15 to 60 seconds. For example, a frequency spectrum can be generated by Fourier transforming data for each 20 seconds with a 5-second shift (15-second overlap). From the perspective of obtaining accurate input data, the ratio of the first predetermined time to the second predetermined time (overlap ratio) is preferably 50 to 90%.

[0042] The frequency spectrum, which is input data for the training data TD, is generated by the conversion processing unit 152. The conversion processing unit 152 uses the generated frequency spectrum data as input data, generates training data TD in which the difference in the exercise intensity of the subject during exercise relative to the exercise intensity at AT measured by CPX, which is performed simultaneously with the measurement of the subject's bioelectrical impedance, is used as a correct answer label, and stores the generated training data TD in the memory unit 12.

[0043] That is, the training data TD is training data in which the input data is a frequency spectrum generated by Fourier transforming the bioelectrical impedance measured for a subject during exercise for each time series data, and the correct answer label is the difference between the exercise intensity of the subject corresponding to the frequency spectrum and the exercise intensity at AT measured by CPX performed simultaneously with the measurement of the subject's bioelectrical impedance. The input data may also be the squared amplitude of the frequency spectrum for each frequency.

[0044] The exercise intensity at the AT is expressed as an index of the heart rate (HR), pulse rate, VO2 (oxygen uptake), VE / VO2 (pulmonary ventilation / oxygen uptake), VE / VCO2 (pulmonary ventilation / CO2 excretion), R (gas exchange ratio), or the load of the exercise equipment, preferably expressed as the heart rate (HR). The load of the exercise equipment is, for example, the pedal load of a fitness bike.

[0045] The exercise intensity of a subject during exercise is the same as the index of exercise intensity at AT, and is expressed as heart rate (HR), pulse rate, VO2 (oxygen uptake), VE / VO2 (pulmonary ventilation / oxygen uptake), VE / VCO2 (pulmonary ventilation / CO2 excretion), R (gas exchange ratio), or an index of the load of the exercise equipment, and is preferably expressed as heart rate (HR). Calculation models (learning models) can be generated according to the index of exercise intensity at AT and the exercise intensity of the subject during exercise used in the training data TD.

[0046] The difference between the subject's exercise intensity corresponding to the frequency spectrum and the exercise intensity at AT measured by CPX performed simultaneously with the measurement of the subject's bioelectrical impedance may be the quantitative difference of the subject's exercise intensity relative to the exercise intensity at AT, or may be a binary value indicating whether the subject's exercise intensity is higher or lower than the exercise intensity at AT, but is preferably the quantitative difference of the subject's exercise intensity relative to the exercise intensity at AT. For example, if the exercise intensity at AT is represented by the heart rate (HR) at AT and the subject's exercise intensity during exercise is represented by the heart rate (HR) during exercise, the quantitative difference of the subject's exercise intensity relative to the exercise intensity at AT is represented by the difference ΔHR between the heart rate (HR) at AT and the heart rate (HR) during exercise.

[0047] In the exercise support device 10, the input data can be only frequency spectrum data, but other data may also be included as input data. Other input data include the subject's age, sex, height, weight, body surface area (BSA), BMI (Body Mass Index), temperature, humidity, atmospheric pressure, and elapsed time at the time of measurement, and the load during exercise on the exercise equipment, such as the pedal load on a fitness bike.

[0048] An example of a frequency spectrum obtained by Fourier transform of the bioelectrical impedance measured for a subject is shown in Figure 3. The frequency spectrum is a two-dimensional intensity distribution of the bioelectrical impedance, represented by the horizontal axis as the time axis and the vertical axis as the frequency axis.

[0049] The frequency spectrum shown in Figure 3 is a frequency spectrum obtained by Fourier transforming the data of the subject's bioelectrical impedance, which was measured when the subject was at rest while holding the measuring unit 30 of the impedance measuring device on the handlebars of a fitness bike, when the subject was pedaling the fitness bike, and when the subject had finished pedaling, and which was divided into 20-second segments spaced 5 seconds apart.

[0050] The frequency spectrum shown in Figure 3 shows signals (frequency: 100-200 / min) that correspond to the heart rate during the first 2 minutes of rest, when pedaling a fitness bike at 50 times per minute from 2 to 12 minutes, and when the hands are taken off the handlebars after 10 minutes have elapsed, as well as signals (frequency: 50 / min and 100 / min) that correspond to the pedaling cycle, and a signal derived from breathing (frequency: up to 40 / min).

[0051] Figure 4 shows a configuration diagram of an example of a training data collection system for collecting training data TD for machine learning. A subject exercises on an exercise machine such as a fitness bike while wearing an impedance measuring device for measuring bioelectrical impedance, a heart rate monitor or pulse meter 40 for measuring the subject's heart rate or pulse rate, and a breath gas analyzer 50 for analyzing the subject's breath gas. The exercise machine on which the subject exercises is preferably equipped with substantially the same exercise mechanism as the exercise machine used by the user. This improves the accuracy of predicting the difference between the current exercise intensity and the optimal exercise intensity. For example, if the exercise machine used by the user is a fitness bike, the exercise machine on which the subject exercises is preferably also a fitness bike. The impedance measuring device can be attached to the subject via the measurement unit 30 and impedance meter 20 described above. The heart rate monitor or pulse meter 40 can be a conventional measuring device, such as one attached to the ear. The breath gas analyzer 50 can be an analyzer conventionally used to analyze breath gases such as oxygen concentration and carbon dioxide concentration in CPX.

[0052] Bioelectrical impedance data 121, heart rate or pulse rate data 122, and exhaled gas data 123 such as oxygen concentration and carbon dioxide concentration measured while the subject is exercising on the exercise equipment can be stored in the memory unit 12. Each stored data can be displayed on the display unit 13. From the heart rate or pulse rate data 122 and the exhaled gas data 123, an expert can determine the actual AT value 124 based on the heart rate or pulse rate when the AT was reached, the time when the AT was reached, the load on the exercise equipment such as the weight of the pedals of a fitness bike, and the like. The actual AT value 124 can be stored in the memory unit 12 via the operation unit 14. In this application, the actual AT value 124 determined by an expert using the above method is defined as the exercise intensity at the AT measured by CPX, and is expressed as heart rate (HR), pulse rate, VO2 (oxygen uptake), VE / VO2 (pulmonary ventilation / oxygen uptake), VE / VCO2 (pulmonary ventilation / CO2 excretion), R (gas exchange ratio), or the load of the exercise equipment at the AT, and is preferably expressed as heart rate (HR).

[0053] The exercise intensity (the exercise intensity of the subject during exercise) corresponding to each piece of bioelectrical impedance data of the subject measured simultaneously when measuring the AT actual value 124 can be the exercise intensity at any time in each time segment divided into short-term time-series data or the average exercise intensity for each second predetermined time, but is preferably the average exercise intensity for each second predetermined time. The exercise intensity corresponding to each piece of bioelectrical impedance data of the subject can be stored in the memory unit 12.

[0054] The processing unit 15 calculates the difference 125 in exercise intensity based on the measured AT value 124 stored in the storage unit 12 and the exercise intensity corresponding to each bioelectrical impedance data of the subject.

[0055] As described above, bioelectrical impedance data and the exercise intensity difference 125 corresponding to each bioelectrical impedance data with respect to the exercise intensity at the AT (AT actual measurement value 124) are collected as training data TD. The exercise intensity difference 125 is a quantitative difference in exercise intensity corresponding to each bioelectrical impedance data with respect to the exercise intensity at the AT, or a binary value indicating whether the subject's exercise intensity is higher or lower than the exercise intensity at the AT. For example, it is a quantitative value such as the difference ΔHR between the heart rate (HR) at the AT and the heart rate (HR) during exercise, or a binary value where 0 is when the subject's exercise intensity is lower than the exercise intensity at the AT and 1 is when the subject's exercise intensity is equal to or higher than the exercise intensity at the AT. The quantitative difference in exercise intensity is normalized and used as training data TD. For example, normalization can be performed by linear transformation so that the average value and standard deviation of the quantitative difference in exercise intensity for all subjects at all times are both 1.

[0056] (Machine Learning) The following describes the learning process performed by learning processing unit 153. Learning processing unit 153 generates a trained calculation model M1 trained using teacher data TD stored in storage unit 12, and stores the generated trained calculation model M1 in storage unit 12.

[0057] A configuration diagram of an example of a machine learning system when performing machine learning is shown in Figure 5. During machine learning, measured bioelectrical impedance data 121 and exercise intensity difference 125 stored in memory unit 12 are used as training data TD.

[0058] A learning device composed of a conversion processing unit 152 and a learning processing unit 153 preprocesses bioelectrical impedance data 121, and trains a calculation model using training data TD having the preprocessed bioelectrical impedance data as input data and the exercise intensity difference 125 as an output label. The preprocessing is a Fourier transform process of the bioelectrical impedance data. A calculation model M1 can be generated by updating the weighting variables of the learning model so that the difference in exercise intensity relative to the AT calculated (estimated) based on the input bioelectrical impedance data is reduced or matched to the exercise intensity difference 125 that is the output label. To generate the calculation model M1 by machine learning, learning methods such as backpropagation, feedback alignment, direct feedback alignment, synthetic gradient, target prop, difference target prop, and bootstrap can be used.

[0059] If the exercise intensity difference 125 is a quantitative difference in exercise intensity corresponding to each bioelectrical impedance data with respect to the exercise intensity at the AT, the weighting variables of the learning model can be updated to generate a calculation model M1 so that the error of the exercise intensity at the AT calculated (estimated) based on the input bioelectrical impedance data is reduced for the output label, which is the quantitative difference value. For example, if the exercise intensity difference 125 is the difference ΔHR between the heart rate (HR) at the actually measured AT and the heart rate (HR) during exercise, the weighting variables of the learning model can be updated to generate a calculation model M1 so that the error of the difference ΔHR (estimated) between the heart rate (HR) at the AT estimated based on the input bioelectrical impedance data and the heart rate (HR) during exercise is reduced relative to ΔHR (actual measurement).

[0060] When the difference 125 in exercise intensity is a binary value of low or high, where the exercise intensity corresponding to each bioelectrical impedance data is low relative to the exercise intensity at the AT, the weighting variables of the learning model can be updated to generate a calculation model M1 so that the exercise intensity at the AT calculated (estimated) based on the input bioelectrical impedance data matches the output label, where low is 0 and high is 1.

[0061] Figure 6 shows a schematic diagram of an example of a machine learning algorithm. The subject's bioelectrical impedance is acquired as one-dimensional time-series data, for example, at 128 times per second, as data for each time point. The subject's bioelectrical impedance is measured the same number of times as the number of subjects.

[0062] The measured time-series bioelectrical impedance data is divided into segments every short time, for example, every 20 seconds. For example, the first segment can be 0-20 seconds, the next segment 5-25 seconds, and the next segment 10-30 seconds, with a predetermined time lag of, for example, 5 seconds (with a 15-second overlap).

[0063] The segmented bioelectrical impedance data is Fourier transformed and converted into a frequency spectrum. The Fourier transformed frequency spectrum can be normalized to obtain a normalized frequency spectrum. Normalization can be performed, for example, by calculating the average value of the frequency spectra of all subjects and dividing the frequency spectrum of each individual at that frequency by the average value at that frequency. If other input data other than the frequency spectrum is included, normalization is performed for each type of other input data. For example, normalization can be performed using a linear transformation so that the average value and standard deviation of the input data at all times for all subjects are both 1.

[0064] The normalized frequency spectrum and the exercise intensity difference 125 are input into the learning model as training data TD, and the weighting variables of the learning model are updated to reduce or match the error of the output (estimated) exercise intensity difference relative to the AT (hereinafter also referred to as the AT exercise intensity difference (estimate)) relative to the output label, exercise intensity difference 125, thereby generating a calculation model M1.

[0065] Fig. 7 is a schematic diagram showing an example of a calculation model M1. The calculation model M1 shown in Fig. 7 is a neural network model having an input layer M1-1, a hidden layer M1-2, and an output layer M1-3. The frequency spectrum of the training data TD is input to each of the so-called "neurons" of the input layer M1-1. The number of neurons in the hidden layer M1-2 may be more or less than the number of neurons in the input layer M1-1.

[0066] Fig. 8 is a schematic diagram showing another example of the calculation model M1. The calculation model M1 shown in Fig. 8 is a deep neural network model including a convolutional neural network having an input layer M1-1, a convolutional layer M1-2, a pooling layer M1-3, and an output layer M1-4. There may be two or more convolutional layers M1-2 and two or more pooling layers M1-3. The frequency spectrum of the training data TD is input to each of the neurons in the input layer M1-1.

[0067] The learning processing unit 153 generates or updates a calculation model in which the weights of each neuron in the neural network are learned by performing known machine learning using the training data TD. The learning processing unit 153 may also generate or update a calculation model in which the weights of each neuron in the multi-layered neural network are learned by performing known deep learning using the training data TD.

[0068] 9 shows a configuration diagram of an example of the support system when a user uses the exercise support device 10. A predictor made up of a conversion processing unit 152, a learning processing unit 153, and an output processing unit 155 preprocesses the bioelectrical impedance data measured by the impedance measuring device, and uses the preprocessed bioelectrical impedance data as input data to calculate the difference between the current exercise intensity of the user exercising and the optimal exercise intensity using a trained calculation model M1.

[0069] The conversion processing unit 152 performs preprocessing to generate a frequency spectrum to be used in the calculation process by Fourier transforming the acquired bioelectrical impedance data. The conversion processing unit 152 may generate a frequency spectrum to be used in the calculation process by Fourier transforming the acquired bioelectrical impedance data for each time series data.

[0070] The calculation processing unit 154 executes a calculation process. Using the generated frequency spectrum data as input data and the trained calculation model M1, the calculation processing unit 154 calculates the difference between the optimal exercise intensity of the user currently exercising and the current exercise intensity. The difference between the optimal exercise intensity of the user currently exercising and the current exercise intensity can be the ratio of the current exercise intensity to the optimal exercise intensity or the qualitative difference between the optimal exercise intensity and the current exercise intensity.

[0071] The percentage of the current exercise intensity relative to the optimal exercise intensity is the percentage (%) of the current exercise intensity relative to the optimal exercise intensity, indicating how much lower, higher, or equal the current exercise intensity of the user during exercise is relative to the optimal exercise intensity. The percentage of the current exercise intensity relative to the optimal exercise intensity is expressed, for example, as a percentage of the current exercise intensity of the user during exercise, such as 80% or 115% of the optimal exercise intensity, or as a percentage of 20% lower or 15% higher than the optimal exercise intensity.

[0072] The qualitative difference of the current exercise intensity from the optimal exercise intensity is a qualitative difference in which the current exercise intensity of the user during exercise is the optimal exercise intensity, lower than the optimal exercise intensity, or higher than the optimal exercise intensity.

[0073] The optimal exercise intensity of a user during exercise can be set arbitrarily according to the subject's exercise purpose, for example, within a predetermined range around the AT, such as AT ±10%, a predetermined range below or below the AT, such as AT -10% or more and below the AT, or a predetermined range above or above the AT, such as AT or more and AT +10%. The optimal exercise intensity of a user during exercise is preferably within a predetermined range below or below the AT. The predetermined range below or below the AT is preferably within the range below or below the AT and AT -5%, AT -10%, AT -15%, or AT -20%. The user's AT is the user's AT estimated by the exercise support device 10.

[0074] The calculation processing unit 154 transmits the calculation result to the output processing unit 155. The output processing unit 155 outputs information on the calculation result. The output of the information is to display it on the display unit 13 and / or to transmit it to another device, etc.

[0075] 10 is a diagram showing an example of the operation flow of the calculation process by the conversion processing unit 152, the calculation processing unit 154, and the output processing unit 155. The calculation process is executed by the conversion processing unit 152, the calculation processing unit 154, and the output processing unit 155. In the calculation process, when data of the user's bioelectrical impedance is acquired from the impedance measuring device, the "difference between the current exercise intensity of the user during exercise and the optimal exercise intensity of the user" corresponding to the acquired bioelectrical impedance is calculated.

[0076] When bioelectrical impedance data is acquired from the impedance measuring device, the conversion processing unit 152 generates a frequency spectrum by Fourier transforming the acquired bioelectrical impedance data (step S101).

[0077] Next, the calculation processing unit 154 executes a calculation process of the "difference between the current exercise intensity of the user during exercise and the optimal exercise intensity" corresponding to the obtained bioelectrical impedance (step S102).

[0078] The output processing unit 155 outputs information indicating "the difference between the current exercise intensity and the optimal exercise intensity for the user during exercise" (step S103).

[0079] (ΔHR: Estimated difference between AT heart rate and current heart rate) Figure 11 shows an example of measurements in which the exercise support device 10 estimated the difference between the AT heart rate and the current heart rate for 15 users using a neural network as a trained calculation model. The open square plots are data showing the difference between the AT heart rate and the exercise heart rate, which was estimated from bioelectrical impedance using the exercise support device 10 mounted on a fitness bike. The filled circle plots are data showing the difference between the AT heart rate and the exercise heart rate, as determined by an expert based on CPX measurements taken at the same time as measuring bioelectrical impedance.

[0080] In the machine learning to generate the trained calculation model, the results of the short-time Fourier transform (0-180 / min) of the bioelectrical impedance measured for 15 subjects while they were pedaling a fitness bike were used as input values, and the relative difference in exercise intensity to the AT determined by the CPX measured at the same time was used as the output label.

[0081] In the graph in Figure 11, areas below the central 0 bpm line indicate exercise intensity lower than the optimal exercise intensity, while areas above the central 0 bpm line indicate exercise intensity higher than the optimal exercise intensity. As the pedal load on the fitness bike gradually increases over time, the exercise intensity is low in the first half, reaches the optimal exercise intensity in the middle, and exceeds the optimal exercise intensity in the second half. The data in Figure 11 shows that the current exercise intensity relative to the optimal exercise intensity estimated by the exercise support device 10 is very close to that obtained by CPX and expert judgment.

[0082] The calculation model in the exercise support device 10 can be a neural network, a random forest, a support vector machine (SVM), or the like.

[0083] (Verification of prediction accuracy) 12 shows the results of 10-fold cross-validation of data estimated by the exercise support device 10. The output label of the training data TD was set to a binary value of 0 when the exercise intensity of the subject was less than the exercise intensity at the AT and 1 when the exercise intensity was equal to or greater than the exercise intensity at the AT, and a neural network calculation model was machine-learned.

[0084] In Figure 12, the solid line represents the training data (AT assessment results by CPX and experts), the dashed line represents the prediction for the training data, and the dashed line represents the prediction for the test data. For example, in case 1, the AT of subject 1 was predicted using the data of subjects 2 to 10 as training data. In case 2, the AT of subject 2 was predicted using the data of subjects 1 and 3 to 10 as training data. The estimated data for six of the ten subjects (subjects 1, 2, 3, 4, 5, and 9) roughly matched the AT assessment results by CPX and experts.

[0085] 13 shows the results of 10-fold cross-validation of data estimated by the exercise support device 10 using a random forest as a learning model. The estimated data for three of the ten subjects (subjects 3, 4, and 8) roughly matched the AT assessment results by CPX and experts.

[0086] Figure 11 shows a graph of ΔHR (estimated difference between heart rate at AT and current heart rate) estimated by exercise support device 10, but by using the same bioelectrical impedance data as in Figure 11 as input data for the training data and other indices of HR (heart rate), VO2 (oxygen intake), VE / VO2 (pulmonary ventilation / oxygen intake), VE / VCO2 (pulmonary ventilation / CO2 excretion), R (gas exchange ratio), or AT (0, 1 judgment) measured by CPX as output labels and training the calculation model, it is possible to similarly estimate these other indices.

[0087] 15 to 20 show graphs estimating the current HR (heart rate), VO2 (oxygen uptake), VE / VO2 (pulmonary ventilation / oxygen uptake), VE / VCO2 (pulmonary ventilation / CO2 output), R (gas exchange ratio), and AT (0, 1 determination) for 15 users using trained calculation models generated with output labels of HR (heart rate), VO2 (oxygen uptake), VE / VO2 (pulmonary ventilation / oxygen uptake), VE / VCO2 (pulmonary ventilation / CO2 output), and R (gas exchange ratio) measured by CPX. The input data is the same bioelectrical impedance data as in FIG. 11.

[0088] The open square plots represent data indicating the current HR (heart rate), VO2 (oxygen uptake), VE / VO2 (pulmonary ventilation / oxygen uptake), VE / VCO2 (pulmonary ventilation / CO2 excretion), R (gas exchange ratio), and AT (0, 1 determination), estimated from bioelectrical impedance.

[0089] The black circle plots represent data showing HR (heart rate), VO2 (oxygen uptake), VE / VO2 (pulmonary ventilation / oxygen uptake), VE / VCO2 (pulmonary ventilation / CO2 excretion), R (gas exchange ratio), and AT (0, 1 determination) based on CPX measurements and heart rate measurements taken simultaneously with bioelectrical impedance measurements.

[0090] The open square plots and black circle plots shown in Figures 15 to 20 roughly match, and it can be seen that HR (heart rate), VO2 (oxygen uptake), VE / VO2 (pulmonary ventilation / oxygen uptake), VE / VCO2 (pulmonary ventilation / CO2 excretion), R (gas exchange ratio), and AT (0, 1 determination) can be estimated with high accuracy. [Explanation of symbols]

[0091] 10 Exercise support device 11 Communications Department 12 Storage section 121 Bioelectrical Impedance Data 122 Heart rate or pulse rate data 123 Exhaled gas data 124 AT actual measurement value 125 Difference between optimal exercise intensity and current exercise intensity 13 Display section 14 Control section 15 Processing section 151 Receiving processing unit 152 Conversion processing section 153 Learning processing unit 154 Discrimination processing unit 155 Output Processing Unit 20 Impedance meter 30 Measuring part 31 Current application electrode 32 Voltage measurement electrodes 33 Round bar 40 Heart rate monitor or pulse monitor 50 Exhaled gas analyzer M1 calculation model TD teacher data

Claims

1. An exercise support device that supports a user in exercising at an optimal exercise intensity, an acquisition unit that acquires bioelectrical impedance of the user during exercise; a transform unit that performs a Fourier transform on the acquired bioelectrical impedance to generate a frequency spectrum; a calculation unit that receives the generated frequency spectrum and calculates a difference between the current exercise intensity of the user and an optimal exercise intensity of the user; an output unit that outputs the difference between the calculated optimal exercise intensity and the current exercise intensity; Equipped with the calculation unit has a trained calculation model that has been subjected to machine learning using training data so as to calculate, when the frequency spectrum is input, a difference between the current exercise intensity of the user and an optimal exercise intensity during exercise; The learned calculation model is The input data is a frequency spectrum generated by Fourier transforming the bioelectrical impedance measured for the subject during exercise for each time series data; and The difference in exercise intensity of the subject during the exercise relative to the exercise intensity at the anaerobic threshold measured by a cardiopulmonary exercise stress test using expiratory gas analysis, which is performed simultaneously with the measurement of the bioelectrical impedance of the subject, is used as an output label. Learning was done by Exercise support device.

2. The exercise support device according to claim 1 , wherein the difference of the current exercise intensity from the optimal exercise intensity is the difference between the heart rate at the user's anaerobic threshold during the exercise and the user's current heart rate during the exercise.

3. The exercise support device according to claim 1 , further comprising a display unit that displays a difference between the optimum exercise intensity output from the output unit and the current exercise intensity.

4. The exercise support device according to any one of claims 1 to 3; an impedance measuring device for measuring the bioelectrical impedance of the user during exercise; An exercise support system comprising:

5. The exercise support system according to claim 4 , wherein the measurement device comprises a four-terminal probe measurement unit.

6. 6. The exercise support system according to claim 5, wherein the four-terminal probe measurement unit includes two electrodes arranged on each of the left and right handlebars of the exercise bike.

7. a receiving unit that receives a difference between the current exercise intensity and the optimal exercise intensity transmitted from the output unit; and a display unit that displays whether the current exercise intensity of the user during exercise is higher, the same as, or lower than the optimal exercise intensity of the user, based on the difference between the current exercise intensity and the optimal exercise intensity received by the receiving unit. The exercise support system according to any one of claims 4 to 6, comprising:

8. An exercise support method for supporting a user to exercise at an optimal exercise intensity, comprising: The exercise support device used by the user is acquiring bioelectrical impedance of the user during exercise; Fourier transform the acquired bioelectrical impedance to generate a frequency spectrum; Calculating a difference between the current exercise intensity and the optimal exercise intensity of the user during the exercise by inputting the generated frequency spectrum; and outputting a difference between the calculated optimal exercise intensity and the current exercise intensity; Including, the calculating step uses a trained calculation model that has been subjected to machine learning using training data to calculate a difference between the optimal exercise intensity of the user currently exercising and the optimal exercise intensity of the user when the frequency spectrum is input; The learned calculation model is The input data is a frequency spectrum generated by Fourier transforming the bioelectrical impedance measured for the subject during exercise for each time series data; and The difference in exercise intensity of the subject during the exercise relative to the exercise intensity at the anaerobic threshold measured by a cardiopulmonary exercise stress test using expiratory gas analysis, which is performed simultaneously with the measurement of the bioelectrical impedance of the subject, is used as an output label. Learning was done by Exercise support method.

9. An exercise support program that supports a user to exercise at an optimal exercise intensity, A processing unit of the exercise support device used by the user, a Fourier transform of the bioelectrical impedance measured during the user's exercise to generate a frequency spectrum; and When the generated frequency spectrum is input, a calculation process is performed to calculate a difference between the optimal exercise intensity of the user during exercise and the current exercise intensity. [0033] the calculation process uses a trained calculation model that has been subjected to machine learning using training data so as to calculate a difference between the current exercise intensity of the user and an optimal exercise intensity when the frequency spectrum is input; The learned calculation model is The input data is a frequency spectrum generated by Fourier transforming the bioelectrical impedance measured for the subject during exercise for each time series data; and The difference in exercise intensity of the subject during the exercise relative to the exercise intensity at the anaerobic threshold measured by a cardiopulmonary exercise stress test using expiratory gas analysis, which is performed simultaneously with the measurement of the bioelectrical impedance of the subject, is used as an output label. Learning was done by Exercise support program.

Images