Chewing action analysis method

Abstract

Comprises: an acquisition step for acquiring myopotential data by measuring myopotentials from muscles at a plurality of measurement sites on the face of a subject person when the subject person chews; a calculation step for calculating a plurality of feature quantities on the basis of unit data obtained by dividing the myopotential data by unit time; and a determination step for determining, on the basis of the feature amount, which of a tongue action, a cheek action, and an occlusal action is being performed by the subject person at the time when the unit data is acquired.

Description

Technical Field

[0001] The present invention relates to a method for analyzing chewing actions. Background Art

[0002] In recent years, research on chewing has received attention. Chewing is, for example, a series of processes in which a person breaks down food into small pieces to form a paste-like food mass that can be swallowed. As an effect of chewing, its relevance to various elements in life activities has received attention. For example, efficient absorption of nutrients contained in the food mass, maintenance of a good state of the oral cavity environment, effects on mental state and brain function, and prevention of diseases, etc.

[0003] As a method for evaluating chewing, for example, there is a method of measuring the number of chews or the speed, bite force, etc. and evaluating them. In this method, for example, by changing the food to be chewed, it is possible to compare the degree to which the food promotes the chewing action.

[0004] On the other hand, in this method, it is limited to the evaluation of the chewing action, and a sufficient evaluation of the chewing function cannot be performed. The chewing action is not only a simple chewing action but also various actions. By comprehensively evaluating these actions, it is possible to grasp the detailed form of the chewing action and to evaluate in detail the chewing action that varies according to a person's chewing ability and food physical properties.

[0005] The following discloses techniques related to the evaluation of chewing.

[0006] Prior Art Documents

[0007] Patent Documents

[0008] Patent Document 1: Japanese Unexamined Patent Application Publication No. 2012-232065

[0009] Patent Document 2: International Publication No. 2019 / 022259 Summary of the Invention

[0010] Problems to be Solved by the Invention

[0011] However, a method for comprehensively evaluating the chewing action has not been established. In the evaluation of the chewing action, regarding actions other than chewing, it is not clear what actions to measure and analyze in order to appropriately evaluate the chewing action.

[0012] Therefore, the present disclosure provides a method for analyzing chewing actions that can comprehensively evaluate the chewing action.

[0013] Solutions to the Problems

[0014] The present invention provides, for example, the following solutions.

[0015] (1) A method for analyzing chewing movements, comprising the following steps: an acquisition step of measuring the myoelectric potential of muscles at multiple measurement sites on the face of a subject person during chewing and acquiring myoelectric potential data; a calculation step of calculating a plurality of characteristic quantities based on unit data obtained by dividing the myoelectric potential data by unit time; and a determination step of determining, based on the characteristic quantities, which of the tongue movement, cheek movement, and biting movement the subject person is performing when the unit data is acquired.

[0016] (2) The method for analyzing chewing movements according to (1), wherein the multiple measurement sites include the right masseter muscle, the left masseter muscle, and the vicinity of the center of the suprahyoid muscle group.

[0017] (3) The method for analyzing chewing movements according to (1) or (2), wherein in the determination step, it further includes: in the case of the tongue movement being performed, determining whether the following movements are being performed: whether the tongue is moving upward, downward, rightward, or leftward; whether the tongue is being protruded forward; whether the tongue is being pulled back; whether the inner or outer sides of the teeth on the right or left are being wiped, and in the case of the cheek movement being performed, determining whether the molars are being wiped with the left cheek, right cheek, or both cheeks.

[0018] (4) The method for analyzing chewing movements according to any one of (1) to (3), wherein the characteristic quantities are calculated based on the waveform of the myoelectric potential of each measurement site and a waveform data set obtained by aggregating the myoelectric potentials of the multiple measurement sites.

[0019] (5) The method for analyzing chewing movements according to any one of (1) to (4), wherein a part of the data of the first unit data is repeated with the second unit data at the time before the first unit data.

[0020] (6) The method for analyzing chewing movements according to any one of (1) to (5), wherein in the determination step, in the determination of which movement is being performed, the probability of each movement being performed is calculated.

[0021] (7) The method for analyzing chewing movements according to any one of (1) to (6), wherein the determination step is executed by a learned model obtained by learning using the myoelectric potential data of the multiple measurement sites when the tongue movement, the cheek movement, and the biting movement are being performed as teacher data.

[0022] (8) The method for analyzing chewing movements according to any one of (1) to (7), wherein the teacher data includes the myoelectric potential data of the multiple measurement sites during quiet times when chewing is not being performed.

[0023] (9) The chewing motion analysis method according to any one of (1) to (8), wherein the teacher data includes myoelectric potential data of the plurality of measurement sites during swallowing with a swallowing motion.

[0024] (10) The chewing motion analysis method according to any one of (1) to (9), wherein the determination step determines that the motion at rest and the motion during swallowing are not any of the tongue motion, the cheek motion, and the biting motion.

[0025] (11) The chewing motion analysis method according to any one of (1) to (10), wherein the plurality of measurement sites includes the left and right sides of the suprahyoid muscle group.

[0026] (12) The chewing motion analysis method according to any one of (1) to (11), wherein the plurality of measurement sites includes the left temporalis muscle and the right temporalis muscle.

[0027] (13) The chewing motion analysis method according to any one of (1) to (12), wherein the plurality of measurement sites includes the left buccinator muscle and the right buccinator muscle.

[0028] (14) The chewing motion analysis method according to any one of (1) to (13), further comprising a food difference evaluation step, in which, based on the determination results when the subject person chews different foods, the difference in the motion during chewing due to the difference in foods is detected.

[0029] (15) The chewing motion analysis method according to any one of (1) to (14), further comprising a time elapsed evaluation step, in which the change in the motion during chewing that occurs over time is detected.

[0030] Effects of the Invention

[0031] The present disclosure can comprehensively evaluate chewing motions. Description of the Drawings

[0032] Figure 1 is a diagram showing a structural example of the chewing motion analysis system 10.

[0033] Figure 2 is a diagram showing a structural example of the chewing motion analysis device 100.

[0034] Figure 3 is a diagram showing an example of the software structure of the chewing motion analysis system 10 and the input / output data of each part during chewing motion determination.

[0035] Figure 4 This is a diagram showing an example of the software structure of the chewing motion analysis system 10 and the input / output data of each part during model learning.

[0036] Figure 5 This is a diagram showing an example of the myoelectric potential data of each measurement site.

[0037] Figure 6 This is a diagram showing an example of the process flow chart of the feature quantity calculation process S100.

[0038] Figure 7 This is a diagram showing an example of the myoelectric potential data and unit data.

[0039] Figure 8 This is a diagram showing an example of the calculated feature quantities.

[0040] Figure 9 This is a diagram showing an example of the process flow chart of the chewing motion analysis process S200.

[0041] Figure 10 This is a diagram showing an example of the chewing motion to be judged.

[0042] Figure 11 This is a diagram showing an example of the chewing motion judgment result.

[0043] Figure 12 This is a diagram showing an example of the process flow chart of the model learning process S300.

[0044] Figure 13 This is a diagram showing an example of the chewing motion based on different chewing substances.

[0045] Figure 14 This is a diagram showing an example of the change in the chewing motion that occurs over time.

[0046] Figure 15 This is a diagram showing an example of the real-time display of the chewing motion analysis.

[0047] Figure 16 This is a diagram showing an example of the chewing motion including swallowing based on different chewing substances.

[0048] Figure 17 This is a diagram showing an example of the change in the chewing motion including swallowing that occurs over time. Detailed implementation mode

[0049] [First implementation mode]

[0050] The first implementation mode will be described.

[0051] <Structural example of the chewing motion analysis system 10>

[0052] Figure 1 This is a diagram showing a structural example of the chewing motion analysis system 10. The chewing motion analysis system 10 includes a chewing motion analysis device 100 and electromyogram measurement pads P1 to P3. The chewing motion analysis system 10 is a system that measures the electromyogram from the muscles of the face of the measurement subject 20 and determines and analyzes what kind of motion the measurement subject 20 is performing during chewing based on the measured data.

[0053] The electromyogram measurement pads P1 to P3 are pads that are attached to the face of the measurement subject 20 of the chewing motion and are used to measure the electromyogram indicating the amount of motion of the muscles of the face of the measurement subject 20. The electromyogram measurement pads P1 to P3, for example, send the measured electromyogram to the chewing motion analysis device 100 via a cable. In addition, regarding the electromyogram measurement pads, Figure 1 there are three in number here, but the number of settings can be increased according to the measurement site.

[0054] The chewing motion analysis device 100 is a device that acquires the measured electromyogram data and determines and analyzes what kind of motion the measurement subject 20 is performing during chewing based on the measured data. The chewing motion analysis device 100 is, for example, a computer. The chewing motion analysis device 100 outputs the result obtained by determining the motion as data or displays it on a screen.

[0055] <Structural example of the chewing motion analysis device 100>

[0056] Figure 2 This is a diagram showing a structural example of the chewing motion analysis device 100. The chewing motion analysis device 100 includes a CPU (Central Processing Unit) 110, a storage unit 120, a memory 130, a communication circuit 140, and a display 150.

[0057] The storage unit 120 is an auxiliary storage device such as a flash memory, an HDD (Hard Disk Drive), or an SSD (Solid State Drive) that stores programs and data. The storage unit 120 stores an electromyogram measurement program 121, a feature quantity calculation program 122, a chewing motion recognition model 123, and a model learning program 124.

[0058] The chewing motion recognition model 123 is, for example, AI (artificial intelligence), and can determine what motion (chewing motion) is being performed during chewing based on the feature quantities of the measured myoelectric potential data, and output the probabilities of each chewing motion being performed. The chewing motion recognition model 123 uses the learned model when making a determination of the chewing motion. The chewing motion recognition model 123 improves the determination accuracy of the chewing motion by being input with the teacher data for learning. In the following description, unless otherwise specified, the chewing motion recognition model 123 is a learned model that has completed a specified amount (specified number of times) of learning and has a determination accuracy above a specified level.

[0059] The memory 130 is an area for loading the programs stored in the storage unit 120. In addition, the memory 130 can also be used as an area for the programs to store data.

[0060] The communication circuit 140 is a device that receives the signals detected by the myoelectric potential measurement pads P1 to P3 through the receiver included in the myoelectric potential data measurement unit 11 and obtains the myoelectric potential data via a cable.

[0061] The display 150 is a screen for displaying the determination result and analysis result of the motion, and is a display unit. The display 150 can be, for example, an integrated display or an externally connected display.

[0062] The CPU 110 is a processor that loads the programs stored in the storage unit 120 into the memory 130 and executes the loaded programs to construct each unit to implement each process.

[0063] The CPU 110 constructs the myoelectric potential data measurement unit by executing the myoelectric potential measurement program 121 to perform the myoelectric potential data measurement process. The myoelectric potential data measurement process is a process of causing the myoelectric potential measurement pads P1 to P3 to measure the myoelectric potential and send the myoelectric potential data to obtain the myoelectric potential data.

[0064] The CPU 110 constructs the feature quantity calculation unit by executing the feature quantity calculation program 122 to perform the feature quantity calculation process. The feature quantity calculation process is a process of dividing the input myoelectric potential data by a determination unit (time unit) and calculating the feature quantities of each segmented data obtained by the division.

[0065] The CPU 110 constructs the chewing motion recognition unit by executing the chewing motion recognition model 123 to perform the chewing motion recognition process. The chewing motion recognition process is a process of outputting the chewing motion corresponding to the input feature quantity. The chewing motion recognition process, for example, calculates the probabilities of each chewing motion being performed and outputs the probabilities.

[0066] The CPU 110 constructs a model learning unit by executing a model learning program 124 to perform model learning processing. The model learning processing is a process of generating teacher data to be input when the chewing motion recognition model 123 (including a learned model or an unlearned model) is learned. The teacher data is composed of, for example, the feature amounts of myoelectric potential data obtained by measuring a deliberately performed chewing motion and the category of the deliberately performed motion (correct data). The model learning processing is a process of generating or editing teacher data and inputting the generated or edited teacher data into the chewing motion recognition model 123 to cause the chewing motion recognition model 123 to learn.

[0067] <Example of the software structure of the chewing motion analysis system 10>

[0068] The chewing motion analysis system 10 includes a myoelectric potential data measurement unit 11, a feature amount calculation unit 12, a chewing motion recognition unit 13, and a model learning unit 14. Hereinafter, the processing and input / output of each unit at the time of chewing motion determination for determining a chewing motion and at the time of model learning for learning the chewing motion recognition model 123 will be described.

[0069] <1. At the time of chewing motion determination>

[0070] Figure 3 It is a diagram showing an example of the software structure of the chewing motion analysis system 10 and the input / output data of each unit at the time of chewing motion determination.

[0071] The myoelectric potential data measurement unit 11 measures myoelectric potential data (S10) during actual eating (chewing of food) by the determination target person and outputs the measured myoelectric potential data (S11). The myoelectric potential data measurement unit 11 measures myoelectric potential data for a specified time using, for example, myoelectric potential measurement pads P1 to P3.

[0072] The feature amount calculation unit 12 is input with myoelectric potential data (S12) and outputs a feature amount (S13). The feature amount calculation unit 12 divides the myoelectric potential data by time units and calculates the feature amount for each unit data obtained by the division. In addition, the feature amount calculation unit 12 may calculate the feature amount considering the unit data of multiple measurement sites. Further, the feature amount calculation unit 12 may consider the unit data of other times when calculating the feature amount of the unit data of one measurement site.

[0073] The chewing motion recognition unit 13 is input with the feature amount (S14) and outputs, as a determination result, the probability of the chewing motion of each unit data (the probability indicating which chewing motion is being performed at the acquisition time of the unit data) (S15).

[0074] The model learning unit 14 is a processing unit that processes actions during the learning of the chewing motion recognition model 123 and does not specifically process actions during chewing motion determination.

[0075] <2. During model learning>

[0076] Figure 4 It is a diagram showing an example of the software structure of the chewing motion analysis system 10 and the input / output data of each unit during model learning.

[0077] The myoelectric potential data measurement unit 11 measures myoelectric potential data (S20) during the simulated actions for generating teacher data and outputs the measured myoelectric potential data (S21). The simulated actions are, for example, actions that extract the chewing actions, tongue-using actions, and actions of the inner wall of the oral cavity using the cheeks during actual meals. In this case, the myoelectric potential data measurement unit 11 measures myoelectric potential data during actual meals.

[0078] In addition, the simulated actions can also be actions such as intentionally having the subject perform chewing actions to be learned, chewing actions for which the determination accuracy is to be improved, or chewing actions with a low occurrence frequency in normal chewing actions.

[0079] The feature quantity calculation unit 12 is input with the myoelectric potential data (for teacher data) (S22) and outputs the feature quantity (for teacher data) (S23).

[0080] The model learning unit 14 is input with the feature quantity (for teacher data) (S24), edits the feature quantity data, generates correct data indicating what action the chewing action performed through this simulated action is, and outputs the correct data and the edited feature quantity data as teacher data (S25).

[0081] The chewing motion recognition unit 13 is input with the teacher data (S26) and performs the learning of the chewing motion recognition model 123.

[0082] <Chewing motion determination process>

[0083] Each process of the chewing motion determination process for determining chewing motions will be described.

[0084] <1. Myoelectric potential data measurement process>

[0085] The myoelectric potential data measurement process is a process of measuring myoelectric potential from each facial muscle of the measurement subject, and is, for example, executed by the myoelectric potential data measurement unit 11. The myoelectric potential measurement process is, for example, a process of attaching the myoelectric potential measurement pads P1 to P3 to the face of the measurement subject and having the measurement subject chew to measure the myoelectric potential.

[0086] The facial muscles that are measured include, for example, the right masseter muscle (hereinafter sometimes referred to as the right masseter muscle), the left masseter muscle (hereinafter sometimes referred to as the left masseter muscle), and the vicinity of the center of the suprahyoid muscle group (hereinafter sometimes referred to as the center of the suprahyoid muscle group).

[0087] Figure 5 It is a diagram showing an example of myoelectric potential data at each measurement site. Figure 5 Three graphs are shown, showing the measured values ​​of the right masseter muscle, the measured values ​​of the left masseter muscle, and the measured values ​​of the center of the suprahyoid muscle group from the top. In the graph, the vertical axis is the measured myoelectric potential, and the horizontal axis is the measurement time.

[0088] Regarding myoelectric potential, for example, if the myoelectric potential is 0, it means that the muscle is not active. The larger the absolute value of the myoelectric potential, the more active muscle fibers are. Figure 5 In the figure, the longer the time is in the vertical direction, the more active the muscle fibers are in the measurement area.

[0089] <2. Feature Quantity Calculation Processing>

[0090] The feature amount calculation process S100 is a process of calculating a feature amount based on myoelectric potential data, and is executed by the feature amount calculation unit 12 , for example. Figure 6 1 is a diagram showing an example of a process flowchart of the feature amount calculation process S100 .

[0091] The feature quantity calculation unit 12 receives the myoelectric potential data ( S100 - 1 ).

[0092] The feature amount calculation unit 12 cuts the myoelectric potential data into time units of a predetermined time and creates unit data (S100-2).

[0093] Figure 7 It is a diagram showing examples of myoelectric potential data and unit data. Figure 7 The upper part of the figure is Figure 5 A graph obtained by superimposing the measurement data of each part in .

[0094] The feature quantity calculation unit 12 extracts a predetermined time unit T1 from the myoelectric potential data to generate unit data D1. Next, the feature quantity calculation unit 12 extracts the predetermined time unit T1 by shifting the time by the amount T2 from the head time of the unit data D1 as the head time to generate unit data D2. The feature quantity calculation unit 12 repeatedly extracts the myoelectric potential data of the predetermined time unit T1 by shifting the amount T2 until unit data Dn (n is an integer) is generated.

[0095] In the feature quantity calculation process S100, by setting the shift amount and making a part of the time of the unit data repeat, the waveform of the myoelectric potential data can be used without omission in the determination of the chewing motion. In addition, in the feature quantity calculation process S100, by making a part of the time of the unit data repeat, a feature quantity considering the tendency (change tendency) of the waveform of the myoelectric potential data can be calculated.

[0096] In addition, the feature quantity calculation unit 12 generates unit data for each measurement site. In this case, the feature quantity calculation unit 12 calculates the feature quantity for each measurement site. In addition, the feature quantity calculation unit 12 may generate unit data based on the waveform data group obtained by aggregating the unit data and the composite wave obtained by synthesizing the myoelectric potential data of multiple measurement sites. In this case, the feature quantity calculation unit 12 calculates the feature quantity reflecting the measurement results of multiple measurement sites. Also, the feature quantity calculation unit 12 may generate unit data for either the waveform data group obtained by aggregating the unit data or the composite wave for each measurement site. In this case, the feature quantity calculation unit 12 calculates both the feature quantity for each measurement site and the feature quantity reflecting the measurement results of multiple measurement sites. In this case, by determining the chewing motion based on the feature quantity considering both the individual tendency of the measurement site and the tendency of the entire facial muscle, compared with the case of determining only based on individual feature quantities, only based on the feature quantity of the waveform data obtained by aggregating the unit data, or only based on the synthesized feature quantity, sometimes a higher-precision determination can be made.

[0097] Return to Figure 6 , the feature quantity calculation unit 12 calculates the feature quantity for each unit data (S100-3), and ends the process.

[0098] Figure 8 It is a diagram showing an example of the calculated feature quantity. Examples of the calculated feature quantity include root mean square, absolute value average, integral value, number of zero crossings, number of extreme values, wavelength, median frequency, average frequency, autoregressive model nth coefficient (n is an integer from 1 to 6), etc.

[0099] The root mean square is, for example, the square root of the arithmetic mean of the values obtained by squaring the measured values. The root mean square represents, for example, a feature quantity related to the amplitude of the waveform of the measurement data.

[0100] The absolute value average is, for example, the average of the absolute values of the measured values. The absolute value average represents, for example, a feature quantity related to the amplitude of the waveform of the measurement data.

[0101] The integral value is, for example, the interval area calculated by multiplying the measured values between unit data by the sampling time of the myoelectric potential and adding them all up. The integral value represents, for example, a feature quantity related to the amplitude of the waveform of the measurement data.

[0102] The number of zero crossings is, for example, the number of times the measured data changes from a value other than 0 to 0 (counted as 1 time when 0 is continuous). The number of zero crossings represents, for example, a characteristic quantity related to the frequency of the waveform of the measured data.

[0103] The number of extreme values is, for example, the number of times the measured data becomes the highest (peak vertex) within a series of times. The number of extreme values represents, for example, a characteristic quantity related to the frequency of the waveform of the measured data.

[0104] The wavelength is, for example, the wavelength of the waveform of the measured data. The wavelength represents, for example, a characteristic quantity related to the frequency of the waveform of the measured data.

[0105] The median frequency is, for example, the frequency that divides the area in half in the power spectrum of the frequencies obtained by performing a fast Fourier transform on the measured data. The median frequency represents, for example, a characteristic quantity related to the frequency of the waveform of the measured data.

[0106] The average frequency is, for example, the average value of each frequency in the power spectrum of the frequencies obtained by performing a fast Fourier transform on the measured data. The average frequency represents, for example, a characteristic quantity related to the frequency of the waveform of the measured data.

[0107] An autoregressive model is, for example, a model that uses the outputs at times earlier than the output at a certain time to perform regression on the output at that certain time. The autoregressive model represents, for example, a characteristic quantity related to the tendency of the waveform of the measured data. The autoregressive model is represented by the following formula (1).

[0108] X t =a 1 X t-1 +a 2 X t-2 +…+a n X t-n … Formula (1)

[0109] X t is the output at time t, and a n is the coefficient related to the output at time n, called the nth coefficient. The feature quantity calculation unit 12 calculates, for example, the 1st coefficient to the 6th coefficient as feature quantities.

[0110] In the chewing motion analysis system 10, the feature quantity calculation unit 12 calculates a large number of feature quantities as described above and inputs them to the chewing motion recognition unit 13, thereby enabling a higher-precision analysis of the chewing motion.

[0111] <3. Chewing Motion Analysis Process>

[0112] The chewing motion analysis process S200 is a process of calculating feature quantities based on the myoelectric potential data, and is, for example, executed by the chewing motion recognition unit 13. Figure 9It is a diagram showing an example of a process flow chart of the chewing motion analysis process S200.

[0113] The chewing motion recognition unit 13 is input with feature quantities (S200-1).

[0114] Based on the input feature quantities, the chewing motion recognition unit 13 analyzes the feature quantities using an analysis method for chewing motions (S200-2).

[0115] The analysis method for chewing motions will be described. The analysis method for chewing motions is, for example, an analysis method that estimates using a decision tree constructed through the learning of the chewing motion recognition model 123. The generation of the decision tree is described in the model learning process.

[0116] The chewing motion recognition unit 13 uses the decision tree to branch out each feature quantity and calculates the corresponding motion (or the probability of the motion).

[0117] The chewing motion recognition unit 13 calculates the probability of each chewing motion being performed (S200-3), and ends the process.

[0118] Figure 10 It is a diagram showing an example of the chewing motions being judged. The chewing motion recognition unit 13, for example, calculates the probability of each of 15 motions being performed.

[0119] The motions being judged are classified into four categories in the major classification. The first category is the biting motion, which is, for example, a motion indicating how the teeth are moving up and down (whether chewing is in progress). As the biting motion, for example, there is the chewing motion.

[0120] The second category is the cheek motion, which, for example, indicates how the cheeks are moving. As the cheek motion, for example, there are motions such as wiping the molars with both cheeks, wiping the molars with the right cheek, and wiping the molars with the left cheek.

[0121] The third category is the tongue motion, which, for example, indicates how the tongue is moving. As the tongue motion, for example, there are motions such as upward, downward, forward, pulling back, right, left, wiping the right teeth (inner side, occlusal surface), wiping the right teeth (outer side), wiping the left teeth (inner side, occlusal surface), and wiping the left teeth (outer side).

[0122] The fourth category is other motions, which, for example, indicates a state of doing nothing. As other motions, for example, there are motions such as being quiet.

[0123] Figure 11 It is a diagram showing an example of the chewing motion determination result. Figure 11 Shows the probability of the chewing motion being in progress.

[0124] Figure 11The figure (A) shows an example of a graph representing the probability of the chewing motion before smoothing. In the graph, the vertical axis represents the probability of the chewing motion, and the horizontal axis represents time (unit data).

[0125] Figure 11 The figure (B) shows an example of a graph representing the probability of the chewing motion after smoothing. Smoothing is performed, for example, by generating a moving average of multiple (e.g., three) unit data instead of generating a single unit data. It can be seen that in Figure 11 the figure (B), the change in the waveform with a high probability of the chewing motion becomes smoother compared to Figure 11 the figure (A). In actual chewing, for example, sudden changes in the chewing motion such as suddenly stopping or starting the chewing motion are less frequent, and most often the motion gradually changes from one motion to another. Therefore, by using the smoothed data like this, the change in the probability of the chewing motion calculated during a series of chewing becomes smoother, and the chewing motion can be estimated more accurately.

[0126] <4. Model Learning Process>

[0127] The model learning process S300 is a process of making the chewing motion recognition model 123 learn, and is, for example, executed by the model learning unit 14. Figure 12 The figure shows an example of a flowchart of the model learning process S300.

[0128] The model learning unit 14, for example, causes the myoelectric potential data measurement unit 11 to measure the myoelectric potential data during the simulated motion (S300-1).

[0129] Next, the model learning unit 14, for example, causes the feature quantity calculation unit 12 to execute the feature quantity calculation process and outputs the feature quantity of the myoelectric potential data (S300-2).

[0130] Next, the model learning unit 14 combines the feature quantity with the correct data (data indicating what motion the chewing motion is during the simulated motion) to create teacher data (S300-3).

[0131] Then, the model learning unit 14 inputs the teacher data into the chewing motion recognition model 123, causes the chewing motion recognition model 123 to perform learning (S300-4), and ends the process.

[0132] The chewing motion recognition model 123 recognizes the relationship (tendency) between the feature quantity and the chewing motion performed when presenting the feature quantity based on the teacher data, and performs learning.

[0133] Regarding the myoelectric potential data and chewing movements, their correlation was confirmed. For example, regarding the chewing movement, in most cases, the activity level of the left and right masseter muscles is much greater than that of the suprahyoid muscles. On the other hand, regarding movements that cause the tongue to move upward, such as the movement of wiping the outside with the tongue, in most cases, the activity of the suprahyoid muscles is much greater than that of the left and right masseter muscles. Additionally, regarding movements that use the cheeks, such as wiping with the cheeks, in most cases, the difference in activity between the left and right masseter muscles and the suprahyoid muscles is less compared to the above-mentioned chewing movements and tongue movement actions.

[0134] The chewing movement recognition model 123, for example, learns the tendency of the relationship between the myoelectric potential data and chewing movements as described above based on feature quantities, and estimates the chewing movement.

[0135] During learning, the chewing movement recognition model 123, for example, generates branches of a decision tree to be used when estimating the chewing movement based on feature quantities, or generates thresholds for the branches. For example, when the chewing movement recognition model 123 is a learned model, the chewing movement recognition model 123, for example, improves the estimation accuracy of the chewing movement by changing the branches (including addition and deletion), or changing (updating) the thresholds of the branches.

[0136] In addition, the chewing movement recognition model 123 can also be a model other than a decision tree. For example, it can be a neural network or other AI. That is, the chewing movement recognition model 123 only needs to be an AI that can learn the relationship between the myoelectric potential data obtained by measuring various parts of the face and the chewing movement as teacher data, and estimate the chewing movement based on the myoelectric potential data of various parts of the face.

[0137] In the model learning process S300, for example, all 15 chewing movements are performed as simulated movements to measure the myoelectric potential. Additionally, when performing the simulated movements, the speed and rhythm of the movements can also be changed to measure the myoelectric potential.

[0138] In the model learning process S300, in order to calculate the feature quantities when transitioning from one chewing movement to a different chewing movement, two or more chewing movements can also be combined, and the combined chewing movements are performed in sequence to measure the myoelectric potential. And in the model learning process S300, for example, mathematical processing can also be applied to the calculated feature quantities, or the tendency in the feature quantities can be emphasized, or new feature quantities can be generated as teacher data.

[0139] In addition, the model learning process S300 can also use the measured data of the actual determination as teacher data in addition to the simulated movements. In this case, during the measurement, it is necessary to additionally record the relationship between the chewing movement and time (which chewing movement was performed at which time).

[0140] In addition, as teacher data, measurement data of each measurement site at rest (when not chewing) can also be used. Even when the subject is not chewing, the muscles are slightly active, such as during breathing and blinking. Therefore, by using the measurement data at rest as teacher data, for example, it is possible to prevent misidentifying electromyogram data with a small activity level at rest as a chewing action. In addition, by using the measurement data at rest as teacher data, for example, the measurement data at rest can be removed as stable noise data, thereby improving the determination accuracy of the chewing action.

[0141] <Evaluation using the determination result of the chewing action>

[0142] The result obtained by determining the chewing action can be used in various evaluations. Below, the evaluation will be described.

[0143] <1. Evaluation of the chewing action based on different chewing substances>

[0144] Figure 13 It is a diagram showing an example of the chewing action based on different chewing substances. Figure 13 It shows the chewing action and its ratio (percentage) of Food A and Food B. In addition, in Figure 13 Regarding the chewing action, the chewing action with the highest probability in its unit data is selected.

[0145] According to Figure 13 Regarding Food B, compared to Food A, there are more movements of the cheeks during chewing. For example, if it is said that moving the cheeks also promotes the activity of the facial muscles, then Food B is assumed to be advertised and sold as a "food for preventing cheek sagging".

[0146] In addition, according to Figure 13 Regarding Food A, compared to Food B, there are more movements of the tongue moving outside the teeth. For example, if it is said that by moving the tongue significantly outside, the activity level of the tongue is increased, thereby having the effect of promoting saliva production, then Food A is assumed to be advertised and sold as a "food for increasing saliva production".

[0147] In this way, by detecting the differences in the chewing actions of each food and evaluating them, it is possible to identify the food that promotes the movement of which part (tongue, teeth, cheeks, etc.), and thus the food can be advertised and sold as a food with various effects.

[0148] In addition, for example, in rehabilitation in medicine, etc., it is possible to estimate the activity level of each part, or it is possible to propose or provide the subject with food corresponding to the muscles for which the maximum output value is particularly desired to be increased or the muscles for which the motor function is desired to be improved.

[0149] In this way, by comparing the differences in chewing actions of multiple foods, the characteristics of the foods can be understood, and thus foods corresponding to the needs can be presented.

[0150] <2. Evaluation of changes in chewing actions over time>

[0151] Figure 14 It is a diagram showing an example of changes in chewing actions over time. Figure 14 It shows the chewing actions and their proportions (percentages) over time for a certain food. In addition, in Figure 14 , regarding the chewing actions, the chewing action with the highest probability in its unit data is selected. Also, the "initial stage", "mid-stage", and "late stage" of the time elapsed are, for example, the times obtained by dividing the overall time from the start of chewing the food until just before swallowing the food into three equal parts. For example, when the overall time is 15 seconds, the "initial stage", "mid-stage", and "late stage" are 5 seconds each.

[0152] According to Figure 14 the following example can be known: When chewing a certain food, the chewing action is the most frequent in the mid-stage, and the occurrence frequency of the chewing action is less in the initial stage and the late stage.

[0153] In addition, according to Figure 14 , the action of the subject person moving the tongue to the right during chewing decreases over time, but the action of moving the tongue to the left increases over time. Based on this, it can be speculated that during the chewing action, the subject person chews while moving the chewing material left and right, rather than always chewing the chewing material against only one side of the oral cavity.

[0154] In this way, by detecting and evaluating the differences in chewing actions over time, the characteristics of the person's chewing actions can be captured. In addition, by comparing together with the differences in chewing actions caused by food differences, the comprehensive differences between the differences caused by food and the differences caused by time elapsed can be detected (for example, regarding Food A, compared to Food B, the tongue movement is more active even over time), and the relationship between the characteristics of the food and the chewing actions of the subject person can also be detected.

[0155] <Display of chewing actions>

[0156] Figure 15 It is a diagram showing an example of real-time display of chewing action analysis. Figure 15 For example, it is displayed on the display 150 of the chewing action analysis device 100.

[0157] Real-time display, for example, can analyze while measuring myoelectric potential data during chewing movements and display the results. Real-time display, for example, shows the data file name and other information (whose data, what food is being chewed, etc.).

[0158] In addition, real-time display, for example, shows a graph G1 of the processing time (elapsed time since the start of chewing), the number of chewing movements (number of chews), the average interval, and the deviation (distribution, standard deviation, etc.).

[0159] In addition, real-time display shows the proportion of 15 chewing movements. Among the chewing movements performed, the chewing movement with the highest probability at that moment is determined. In addition, the proportion is displayed in graph G2 and graph G3. Graph G2, for example, is a graph showing 15 chewing movements respectively. Graph G3, for example, is a graph showing the proportion of large categories.

[0160] In addition, real-time display, for example, shows the history of tongue behavior. The history is displayed in the oral cavity schematic diagram G4. For example, the longer the tongue exists, the darker the color represents the history.

[0161] In addition, real-time display shows the muscle workload of each ch (measurement site). The muscle workload is, for example, normalized by dividing by the amplitude of the myoelectric potential data when the maximum muscle strength is exerted, etc., and shows the degree of muscle strength exerted during the measurement. The muscle workload is, for example, displayed in graph G5. Graph G5, for example, shows the muscle workload of each measurement site based on the passage of time.

[0162] In addition, real-time display shows the diagnosis of the eating style. The diagnosis of the eating style is used, for example, to comprehensively determine the tendencies of the chewing style (number of chews, frequency, rhythm, intensity, etc.) and the eating style (proportion of activities of the tongue, cheeks, occlusion, bias in the movement pattern), and to classify them. For example, in graph G6, the proportion and time passage of each large category of chewing movements are shown, and the results obtained by classification are also displayed.

[0163] In addition, the display content of the real-time display may not be real-time but may be displayed after acquiring the measurement data.

[0164] In addition, in order to implement real-time display, the speed of the determination process of chewing movements can also be increased. For example, in order to speed up the feature quantity calculation process, the feature quantities to be calculated can be appropriately selected to reduce the calculation process. In addition, the calculation process during feature quantity calculation can also be simplified by reducing the significant digits of the measurement data of the myoelectric potential.

[0165] [Second Embodiment]

[0166] A description is given of a second embodiment. In the second embodiment, "swallowing" is added as one of the chewing actions to be determined. "Swallowing" is, for example, an act of swallowing chewed food. In addition, "swallowing" also includes acts of swallowing not only the chewed food itself but also moisture discharged from the chewed food, mixtures of the chewed food and saliva, and the like.

[0167] In addition, "swallowing" is, for example, classified into the major category "others" in the Figure 10 classification. In addition, "swallowing" can also be classified, for example, into a new major category "throat movement" indicating an action that causes the throat to move in the Figure 10 classification.

[0168] Figure 16 is a diagram showing an example of a chewing action including swallowing based on the difference in the masticated material. In addition, Figure 16 is a diagram showing an example in which Figure 13 swallowing is added. In Figure 16 , the parts that are different from Figure 13 are shaded, and the values in Figure 13 are shown in parentheses.

[0169] By adding "swallowing" to the determination of the chewing action, the values of "wiping the right teeth (outer side) with the tongue" and "wiping the left teeth (outer side) with the tongue" are reduced for both Food A and Food B. It is assumed that this is because the action of "swallowing" is an action close to "wiping the right teeth (outer side) with the tongue" and "wiping the left teeth (outer side) with the tongue".

[0170] In the case where "swallowing" is not determined, "swallowing" is determined as "wiping the right teeth (outer side) with the tongue" and "wiping the left teeth (outer side) with the tongue". Since "swallowing" is an action of swallowing something, it is not actually "wiping the right teeth (outer side) with the tongue" and "wiping the left teeth (outer side) with the tongue". Therefore, by determining "swallowing", the determination accuracy of the chewing action is improved, and the actions during "swallowing" can be removed as noise data. This has the same effect as determining "quiet".

[0171] In addition, regarding the value of "wiping the molars with the cheeks", it decreased for Food A, but remained the same for Food B. It is assumed that this is a characteristic caused by, for example, the difference in the food. For example, let Food A be gummy candy and Food B be chewing gum. Chewing gum itself is not swallowed, while gummy candy is swallowed. Therefore, when swallowing gummy candy, which is a solid of a certain size, compared with swallowing only moisture in chewing gum, the action of swallowing is performed with a stronger force, so it is assumed that the muscles of the cheeks need to move more.

[0172] In addition, during the passage of chewing time, the value also changes in the same way as the difference in food. Figure 17 It is a diagram showing an example of changes in chewing actions including swallowing that occur over chewing time. In addition, Figure 17 It is a diagram showing an example with Figure 14 swallowing added. In Figure 17 , the parts that are different from Figure 14 are shaded, and the numerical values in Figure 14 are shown in parentheses. The reason for the change in value is the same as the assumption in the above-mentioned Figure 16 . In addition, in Figure 17 , the "swallowing" in the middle period is 0%, and "swallowing" is not being performed, so it has no effect on the values of other chewing actions.

[0173] In addition, Figure 16 and Figure 17 The data are just an example, and the actions whose values change (decrease) by adding "swallowing" are not limited to "wiping the right teeth (outer side) with the tongue", "wiping the left teeth (outer side) with the tongue", and "wiping the molars with the cheeks", and other actions are also possible. The actions whose values change (decrease) may vary depending on the subject, the target food, the setting conditions of the measuring equipment (including the attachment method of the pad, etc.), etc.

[0174] As described above, by adding "swallowing" to the chewing actions and classifying "swallowing" into "others", the accuracy of the values of the actions of chewing food (biting actions, tongue actions, cheek actions) is improved, and the accuracy of food evaluation is also improved.

[0175] In addition, in order to detect "swallowing", the teacher data also needs to include "swallowing". For example, the experimenter makes the subject intentionally perform chewing actions including "swallowing", measures the myoelectric potential data during swallowing, and uses the myoelectric potential data itself and the characteristic quantities as teacher data for model learning. By using this teacher data for model learning, the learned model can detect the "swallowing" action based on the myoelectric potential data.

[0176] In addition, even if it is not an action intentionally made by the experimenter, when the actions performed by the subject are clear in the measurement of myoelectric potential data in practical applications (for example, in the evaluation of the chewing ability of the subject), the myoelectric potential data in this practical application can also be used as teacher data.

[0177] [Third Embodiment]

[0178] A description is given of the third embodiment. In the third embodiment, the measurement site of the myoelectric potential data is changed (added). In the first embodiment, the myoelectric potential data is measured at three locations on the face (right masseter muscle, left masseter muscle, and the center of the suprahyoid muscle group). In the third embodiment, by increasing the measurement sites, the determination accuracy can be further improved. In addition, it is preferable to measure the myoelectric potential data of as many measurement sites as possible to improve the determination accuracy. However, when considering, for example, the burden on the subject, the ease of measurement, the accuracy of the measured values, etc., it may not always be possible to measure all the measurement sites. Therefore, the measurement sites are classified into measurement groups, and the priority order of measurement, etc. will be described below. In addition, in the following description, the priority order is expressed as "priority order x (x is an integer)". The smaller x is, the higher the priority order. If x is the same, it means the priority order is the same.

[0179] <1. First measurement group>

[0180] The first measurement group is the right masseter muscle, the left masseter muscle, and the center of the suprahyoid muscle group. The first measurement group is the most basic measurement site, and the priority order is 1. Regarding the first measurement group, it is easy to measure (easy to attach the pad), and the accuracy of the measured value is also high. In addition, regarding the first measurement group, there is no interference such as hair at the measurement position, and it can be stably measured regardless of age and gender. Also, regarding the right and left masseter muscles, the differences in the biting action and cheek action can be detected. In addition, regarding the center of the suprahyoid muscle group, the movement of the tongue in the longitudinal and front-back directions, the swallowing movement, and the left-right movement of the tongue can be detected.

[0181] For example, as Figure 5 shown, the absolute value of the amplitude of the myoelectric potential data of the suprahyoid muscle group is a very small value compared to the absolute value of the amplitude of the myoelectric potential data of the right and left masseter muscles. Therefore, for example, when detecting the chewing action only by the myoelectric potential data of the muscles related to the tongue such as the suprahyoid muscle group, the detection accuracy of the chewing action is low, and sometimes the overall accuracy of the chewing evaluation decreases. Therefore, as in the first measurement group, the myoelectric potential data of the suprahyoid muscle group, which is a muscle related to the tongue, and the myoelectric potential data of the right and left masseter muscles are measured, and the chewing action is determined. As a result, the detection accuracy of the chewing action including the chewing movement is improved, and the overall accuracy of the chewing evaluation is also improved.

[0182] <2. Second measurement group>

[0183] The second measurement group is the suprahyoid muscles on the left and right (positions on the left and right sides of the suprahyoid muscles closer to the left and right than the center), and the priority is 2. Regarding the second measurement group, it is easy to measure (easy to attach pads). In addition, by measuring the second measurement group, the movement of the tongue can be detected more accurately. Moreover, by measuring the second measurement group, the movement of the tongue during the action of making the oral cavity negative pressure when performing the action of sucking the cheeks can be detected, thereby detecting the action of sucking the cheeks.

[0184] On the other hand, for example, regarding the movement of the tongue, a certain degree of measurement can be performed at the center of the suprahyoid muscles in the first measurement group. Therefore, the priority of the second measurement group is priority 2.

[0185] <3. The third measurement group>

[0186] The third measurement group is the left and right temporalis muscles (the left temporalis muscle and the right temporalis muscle), and the priority is 3. By measuring the third measurement group, the activity information of various chewing actions (biting actions, tongue actions, cheek actions) can be captured. This is because, through the activity of the temporalis muscle, the mandible is fixed, and the tongue can move up, down, left, and right with the mandible as the fulcrum. In particular, by measuring the third measurement group, the movement of the cheek can be detected more accurately. This is because, in the action of wiping the teeth with the cheek, the opening of the lower jaw and the negative pressure operation in the oral cavity using the tongue are required, and the temporalis muscle is related to these actions.

[0187] On the other hand, for example, regarding the temporalis muscle, it is sometimes difficult to attach pads because of hair, and sometimes it is not easy to measure. Therefore, the priority of the third measurement group is priority 3.

[0188] <4. The fourth measurement group>

[0189] The fourth measurement group is the left and right buccinator muscles (the left buccinator muscle and the right buccinator muscle), and the priority is 4. By measuring the fourth measurement group, the action of wiping the teeth with the cheek can be detected more accurately. In addition, by measuring the fourth measurement group, the action of holding a food mass or assisting the left and right movement of the tongue can be detected.

[0190] On the other hand, regarding the buccinator muscle, for example, sometimes the muscle movement is smaller than that of the masseter muscle, or sometimes it interferes with the measurement position of the masseter muscle, or sometimes the myoelectric potential cannot be accurately obtained according to the situation. When accurate values cannot be obtained, the influence of data loss and error becomes larger, sometimes resulting in a decrease in measurement accuracy. Therefore, the priority of the fourth measurement group is priority 4.

[0191] The tester determines the measurement site according to the above priority order. However, the above priority order is only an example. The measurement site can also be determined, for example, according to the characteristics of the subject (age, gender, state of the measurement position, area, etc.). In addition, the measurement site can also be determined according to the chewing motion to be detected.

[0192] Description of Reference Numerals

[0193] 10: Chewing motion analysis system; 11: Myoelectric potential data measurement unit; 12: Feature quantity calculation unit; 13: Chewing motion recognition unit; 14: Model learning unit; 20: Subject; 100: Chewing motion analysis device; 110: CPU; 120: Storage unit; 121: Myoelectric potential measurement program; 122: Feature quantity calculation program; 123: Chewing motion recognition model; 124: Model learning program; 130: Memory; 140: Wireless communication circuit; 150: Display.

Claims

1. A chewing motion analysis method, comprising the following steps: An acquisition step of measuring the myoelectric potential of muscles at a plurality of measurement sites on the face of a subject person during chewing and acquiring myoelectric potential data; A calculation step of calculating a plurality of characteristic quantities based on unit data obtained by dividing the myoelectric potential data by unit time; And A determination step of determining, based on the characteristic quantities, which of a tongue motion, a cheek motion, and a biting motion the subject person is performing when acquiring the unit data.

2. The chewing motion analysis method according to claim 1, wherein the plurality of measurement sites include the right masseter muscle, the left masseter muscle, and the vicinity of the center of the suprahyoid muscle group.

3. The chewing motion analysis method according to claim 1, wherein in the determination step, it further includes: in the case of performing the tongue motion, determining whether the following motions are being performed: whether the tongue is moving upward, downward, rightward, or leftward; whether the tongue is being protruded forward; whether a tongue retraction motion is being performed; whether the inner or outer sides of the right or left teeth are being wiped; in the case of performing the cheek motion, determining whether the left cheek, the right cheek, or both cheeks are being used to wipe the molars.

4. The chewing motion analysis method according to claim 1, wherein the characteristic quantities are calculated based on the waveform of the myoelectric potential of each of the measurement sites and a waveform data set obtained by summarizing the myoelectric potentials of the plurality of measurement sites.

5. The chewing motion analysis method according to claim 1, wherein a part of the data of the first unit data overlaps with the second unit data at the time before the first unit data.

6. The chewing motion analysis method according to claim 1, wherein in the determination step, in the determination of which motion is being performed, the probability of each motion being performed is calculated.

7. The chewing motion analysis method according to claim 1 or 3, wherein the determination step is executed by a learned model obtained by learning using the myoelectric potential data of the plurality of measurement sites when performing the tongue motion, the cheek motion, and the biting motion as teacher data.

8. The chewing motion analysis method according to claim 7, wherein the teacher data includes the myoelectric potential data of the plurality of measurement sites during quiet times when chewing is not being performed.

9. The chewing motion analysis method according to claim 8, wherein the teacher data includes the myoelectric potential data of the plurality of measurement sites during swallowing when a swallowing motion is being performed.

10. The chewing motion analysis method according to claim 9, wherein the determination step determines that the motions during the quiet times and the motions during swallowing are not any of the tongue motion, the cheek motion, and the biting motion.

11. The chewing motion analysis method according to claim 2, wherein the plurality of measurement sites further include the left and right sides of the suprahyoid muscle group.

12. The chewing motion analysis method according to claim 11, wherein the plurality of measurement sites further include the left temporal muscle and the right temporal muscle.

13. The chewing motion analysis method according to claim 12, wherein, the multiple measurement sites further include the left buccal muscle and the right buccal muscle.

14. The chewing motion analysis method according to claim 1, wherein, it further includes a food difference evaluation process, in which, based on the determination results when the subject chews different foods, the differences in the chewing motions caused by the differences in the foods are detected.

15. The chewing motion analysis method according to claim 1, wherein, it further includes a elapsed time evaluation process, in which the changes in the chewing motions occurring over time are detected.

Images