Method for determining chewing movement, method for determining food, and method for evaluating chewing movement

Abstract

To determine whether mastication is performed consciously or unconsciously.SOLUTION: A computer is made to execute an acquisition process to acquire time series data of an interval of mastication by an object person, an analysis process to calculate the gradient of a regression line by analyzing the fluctuation waveform of the time series data; and a determination process to determine the mastication state of the object person in accordance with whether the gradient shows 1 / f fluctuation.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] The present invention relates to a chewing motion determination method, a food determination method, and a chewing motion evaluation method.

Background Art

[0002] In recent years, research on chewing has attracted attention. As effects of chewing, for example, stress reduction, memory improvement, and disease prevention are known. In particular, there are also research results indicating that conscious and continuous chewing has a long-term effect on the short-term memory of the elderly.

[0003] Techniques related to the evaluation of chewing are disclosed below.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0005] However, it has been difficult to objectively determine whether chewing is being performed consciously, for example, by self-report.

[0006] Therefore, one disclosure provides a chewing motion determination method, a food determination method, and a chewing motion evaluation method for determining whether chewing is being performed consciously or unconsciously.

Means for Solving the Problems

[0007] The present invention relates to a method for determining chewing movements, which involves having a computer perform the following steps: an acquisition step of acquiring time-series data of the chewing intervals of a target person; an analysis step of analyzing the waveform of fluctuations in the time-series data and calculating the slope of a regression line; and a determination step of determining the chewing state of the target person depending on whether or not the slope exhibits 1 / f fluctuation.

[0008] Furthermore, the present invention relates to a method for determining chewing movements, wherein, in the determination step, if the gradient does not show 1 / f fluctuation, it is determined that the chewing of the subject person is being performed consciously.

[0009] Furthermore, the present invention relates to a chewing motion determination method, the acquisition step comprising: an acquisition step of acquiring image data of the chewing of the target person for a predetermined period of time; a detection step of detecting the face of the target person from the image data and detecting a plurality of facial feature points from the detected face; and a data acquisition step of determining the state of chewing based on the amount of movement of the facial feature points per unit time and acquiring the time-series data.

[0010] Furthermore, the present invention relates to a method for determining masticatory movement, comprising a measurement step of obtaining the amount of activity of the masseter muscle of the subject person by measuring surface electromyography during the subject person's mastication, and a data acquisition step of determining the state of mastication based on the amount of activity and obtaining the time-series data.

[0011] Furthermore, the present invention relates to a method for determining chewing movements in which the determination step determines that the chewing of the subject person is performed unconsciously if the gradient exhibits a 1 / f fluctuation.

[0012] Furthermore, the present invention relates to a food determination method that causes a computer to perform the following steps: an acquisition step of acquiring time-series data of the chewing intervals of a target person chewing food; an analysis step of analyzing the waveform of fluctuations in the time-series data and calculating the slope of a regression line; and a determination step of determining whether the food is chewed consciously or unconsciously, depending on whether the slope exhibits 1 / f fluctuations.

[0013] Furthermore, the present invention relates to a method for evaluating chewing movements, which involves having a computer perform the following steps: an acquisition step of acquiring first time-series data of chewing intervals when a subject chews while having no other tasks besides chewing, and second time-series data of chewing intervals when the subject chews while having other tasks besides chewing; an analysis step of analyzing the fluctuation waveforms of the first time-series data and the second time-series data, and calculating the first slope of the regression line of the first time-series data and the second slope of the regression line of the second time-series data; and an evaluation step of comparing the first slope and the second slope with the slope in 1 / f fluctuation to evaluate the conscious state during chewing. [Effects of the Invention]

[0014] One disclosure can determine whether chewing is performed consciously or unconsciously. [Brief explanation of the drawing]

[0015] [Figure 1] Figure 1 shows an example of a processing flowchart for the chewing movement detection process. [Figure 2] Figure 2 shows an example of a processing flowchart for the chewing interval acquisition process S1001 using image data. [Figure 3] Figure 3 shows an example of image data and facial feature points for a frame in a video. [Figure 4] Figure 4 shows an example of data illustrating the difference in movement between the midpoint of the nose and the mandible. [Figure 5] Figure 5 shows an example of a processing flowchart for the mastication interval acquisition process S1002, which uses surface electromyography of the masseter muscle. [Figure 6] Figure 6 shows an example of data on masseter muscle activity. [Figure 7] Figure 7 shows an example of a processing flowchart for the fluctuation analysis process S1011 using power spectral density. [Figure 8]FIG. 8 is a diagram showing an example of a processing flowchart of the trend removal fluctuation analysis process S1012. [Figure 9] FIG. 9 is a diagram showing an example of a graph of β for each unit time when performing a single task and when performing a dual task. [Figure 10] FIG. 10 is a diagram showing an example of a graph of α for each unit time when performing a single task and when performing a dual task.

Mode for Carrying Out the Invention

[0016] [First Embodiment] The first embodiment will be described.

[0017] <Regarding the chewing motion determination process> The control of body movements is roughly classified into physical control (feedforward motion control by sensory reflex) such as walking and standing control, and cognitive control (feedback motion control of cognitive information such as vision). The motion cycle in physical control is irregular and shows 1 / f fluctuation, and the motion cycle in cognitive control is known to represent a characteristic frequency, be regular, and not show 1 / f fluctuation.

[0018] 1 / f fluctuation is a fluctuation in which the power spectrum is inversely proportional to the frequency f. 1 / f fluctuation appears in nature, for example, in the babbling of a small stream or the chirping of birds.

[0019] The chewing motion determination process in the first embodiment analyzes the fluctuation of the chewing interval in the chewing motion, and determines whether the chewing motion is conscious or unconscious based on whether it shows 1 / f fluctuation.

[0020] The relationship between conscious and unconscious chewing motions and the result of the fluctuation analysis of the chewing interval will be described in the following examples.

[0021] Figure 1 shows an example of a processing flowchart for the chewing movement determination process. The chewing movement determination process S10 is a process that determines whether chewing is being performed consciously or unconsciously, based on data on the chewing intervals of the person.

[0022] The chewing motion determination process S10 is implemented, for example, by a computer (or processor) executing a chewing motion determination program. Hereafter, the chewing motion determination process S10 will be described assuming that a computer performs the chewing motion determination process S10.

[0023] In the chewing motion determination process S10, the computer acquires time-series data of chewing intervals (S100). Process S100 is, for example, a data acquisition step. The chewing interval refers to the time from one chewing action to the next. For example, the chewing interval is the time from the moment the teeth are brought together and biting down most tightly to the next moment when the teeth are brought together and biting down most tightly. Alternatively, the chewing interval may be, for example, the time from the moment the upper and lower teeth are furthest apart to the next moment when the upper and lower teeth are furthest apart. The computer acquires time-series data of the subject's chewing intervals over a period of, for example, several seconds, tens of seconds, or several minutes.

[0024] The chewing interval time-series data acquisition process S100 is implemented, for example, by the following process. Details of each process will be described later.

[0025] • Processing to acquire chewing intervals using image data S1001 • Chewing interval acquisition process S1002 by measuring surface electromyography of the masseter muscle. Next, in the chewing motion determination process S10, the computer performs a fluctuation analysis (S101) which analyzes the waveform of fluctuations in the time-series data of chewing intervals and calculates the slope of the regression line. Process S101 is, for example, an analysis step that performs fluctuation analysis.

[0026] The fluctuation analysis process S101 is implemented, for example, by the following processes. Details of each process will be described later.

[0027] • Fluctuation analysis processing using power spectral density S1011 • Trend removal fluctuation analysis processing S1012 In each fluctuation analysis process, the computer calculates the gradient of the regression line over time.

[0028] Next, the computer determines whether the slope of the calculated regression line exhibits 1 / f fluctuation (S102). Process S102 is, for example, a determination step. If the slope of the calculated regression line does not exhibit 1 / f fluctuation (No in S102), the computer determines that it was a conscious chewing movement (the chewing movement was performed consciously at that time) (S103) and terminates the process. On the other hand, if the slope of the calculated regression line exhibits 1 / f fluctuation (Yes in S102), the computer determines that it was an unconscious chewing movement (the chewing movement was performed unconsciously at that time) (S104) and terminates the process.

[0029] This allows the computer to determine whether the person whose time-series chewing interval data was acquired was chewing consciously or unconsciously at the time the data was acquired.

[0030] <1. Regarding the time-series data acquisition process S100 for chewing intervals> Two examples of the chewing interval time-series data acquisition process S100 will be described. Note that the chewing interval time-series data acquisition process S100 can be any process that can acquire the chewing interval of the target person, and is not limited to the two examples described below.

[0031] <1.1 Chewing interval acquisition process S1001 using image data> The chewing interval acquisition process S1001 using image data will be described. Figure 2 shows an example of a processing flowchart for the chewing interval acquisition process S1001 using image data. The computer captures image data of the subject person chewing (capture process) (S1001-1). Capture is performed, for example, by a camera or video camera. The image data is, for example, video data that can be processed on a frame-by-frame basis. Alternatively, the image data may be, for example, a series of still images that are continuous in time.

[0032] Next, the computer detects a person's face from the image data and extracts facial feature points (detection step) (S1001-2). Facial feature points are characteristic parts (points) of the face, and multiple points are extracted.

[0033] Figure 3 shows image data from a frame in a video and an example of facial feature points. In Figure 3, the points plotted as circles within the face represent examples of facial feature points. Facial feature points include points with features such as eyes, mouth, nose, contours including cheeks and chin, and eyebrows. For example, facial feature point P1 is a feature point that indicates the vicinity of the center of the nose. Also, for example, facial feature point P2 is a feature point that indicates the vicinity of the lower jaw.

[0034] Returning to the processing flowchart in Figure 2, the computer calculates the amount of movement for each facial feature point (S1001-3). For video data, the amount of movement is the movement from one frame to the next. For a series of still images, the amount of movement is the movement from one still image to the next. The amount of movement is calculated, for example, in terms of the number of pixels.

[0035] Next, the computer calculates a value obtained by subtracting the movement of the midpoint of the nose from the movement of the mandible (S1001-4). The mandible is a part that moves during chewing. On the other hand, the midpoint of the nose is a part that hardly moves even when chewing. Therefore, by subtracting the movement of the midpoint of the nose from the movement of the mandible (calculating the difference), the computer can calculate the relative movement of the mandible with respect to the midpoint of the nose as the axis, and estimate the state of chewing. Note that the movement of the midpoint of the nose may be the movement of facial feature point P1, or it may be calculated considering the movement of facial feature point P1 and surrounding facial feature points. The movement of the mandible may be the movement of facial feature point P2, or it may be calculated considering the movement of facial feature point P2 and surrounding feature points.

[0036] Next, the computer detects the peak value of the calculated difference (S1001-5). The peak value is, for example, the maximum (peak) value within a single chewing time, and it can be assumed that it is detected when the gap between the upper and lower teeth is wide. In addition, when calculating the peak value, the computer may set a predetermined threshold to prevent misidentifying movements of small parts other than chewing as the peak value, and values ​​smaller than the threshold may not be considered as the peak value.

[0037] Next, the computer calculates the time interval between peak values ​​(S1001-6) and terminates the process. The time interval between peak values ​​is, for example, interval T1, which can be considered as the duration of one chewing cycle.

[0038] Figure 4 shows an example of data showing the difference in movement between the midpoint of the nose and the mandible. In the graph in Figure 4, the vertical axis represents the amount of movement (e.g., number of pixels), and the horizontal axis represents the number of frames. Period T1 is the interval between peak values, which is the chewing interval. Note that Figure 4 is an example using video data. The computer, for example, converts the number of frames to time to create time-series data (data acquisition process).

[0039] <1.2 Chewing interval acquisition process S1002 by measuring surface electromyography of the masseter muscle> This section describes the process S1002 for acquiring the chewing interval using surface electromyography (EMG) of the masseter muscle. Figure 5 shows an example of the processing flowchart for the chewing interval acquisition process S1002 using surface electromyography of the masseter muscle.

[0040] The computer measures the activity level of the left and right masseter muscles of the subject and acquires time-series EMG (surface electromyography) data (S1002-1). The EMG value (absolute value) increases as the amount of muscle activity increases (stronger force is applied).

[0041] The computer converts the measured left and right EMG values ​​into absolute values ​​(S1002-2). The computer then performs division normalization on the EMG values ​​during the chewing interval, using the maximum value as the baseline (S1002-3).

[0042] The computer extracts EMG signals that represent a predetermined percentage (e.g., 10%) or more of the maximum electromyographic potential as EMG signals during chewing (S1002-4). This is done to remove data from actions other than chewing (noise data).

[0043] The computer sums the left and right EMG values, which have been normalized by division, and extracts a sum value greater than or equal to a predetermined value (e.g., 0.1) (S1002-5). This is done to remove extremely low sum values ​​(noise data).

[0044] The computer then considers the time from one peak value to the next in the extracted aggregate values ​​as one chewing cycle, calculates the time-series data of the chewing interval (S1002-6), and terminates the process.

[0045] Figure 6 shows an example of masseter muscle activity data. In the graph in Figure 6, the vertical axis represents the sum of the left and right EMG values ​​normalized by division, and the horizontal axis represents elapsed time. Figure 6B is an enlarged view of a portion of area A1 in Figure 6A. The computer acquires the period T2 from one peak value to the next as the chewing interval.

[0046] <2. About fluctuation analysis processing S101> Two examples of the fluctuation analysis process S101 will be described. Note that the fluctuation analysis process S101 is not limited to the two examples described below, as it only needs to be able to analyze fluctuations in the time-series data of chewing intervals. Furthermore, the fluctuation analysis may be performed by AI (Artificial Intelligence) to enable execution in a shorter time, for example. The AI ​​used for fluctuation analysis is composed of a neural network model that can directly estimate the fluctuation coefficient from a portion of the time-series data of chewing intervals.

[0047] <2.1 Fluctuation analysis processing S1011 using power spectral density> Figure 7 shows an example of a processing flowchart for the fluctuation analysis process S1011 using power spectral density. The fluctuation analysis process S1011 using power spectral density calculates the power spectral density of time-series data of chewing intervals and evaluates fluctuations based on the slope of the graph.

[0048] The computer performs linear interpolation of the time-series data of chewing intervals (S1011-1). The computer performs a Fast Fourier Transform to calculate the power spectral density (S1011-2). The computer represents the calculated power spectral density on a log-log graph with frequency on the vertical axis (S1011-3). Then, the computer calculates the slope of the regression line using the least squares method for frequencies from 0.01 Hz to 0.3 Hz, for example (S1011-4). The computer obtains the calculated slope as β and terminates the process.

[0049] The computer evaluates β as a feature of the fluctuation waveform. For example, if β is approximated to 0, it indicates white noise; if it is approximated to -1, it indicates 1 / f fluctuation; and if it is approximated to -2, it indicates Brownian noise. If β is approximated to -1, the computer considers the time-series data of chewing intervals to exhibit 1 / f fluctuation and determines that the subject was performing unconscious chewing movements.

[0050] <2.2 Trend Removal Fluctuation Analysis Processing S1012> Figure 8 shows an example of a processing flowchart for the trend-removed fluctuation analysis process S1012. The trend-removed fluctuation analysis process S1012 is a process that removes the trend (e.g., the mean value) and evaluates the fluctuations. The trend-removed fluctuation analysis process S1012 is called, for example, DFA (Detrended Fluctuation Analysis).

[0051] The computer calculates the average value of the chewing interval time and subtracts this calculated average value from all elements of the original chewing interval time series data (S1012-1). The computer calculates the integral value of the data from the 1st to the ith (where i is an integer greater than or equal to 2) and uses this as the ith data (S1012-2), generating a second time series data. The computer then divides the second time series data into sections of length n (where n is an integer greater than or equal to 4) and calculates a regression line for each section using the least squares method (S1012-3). The computer calculates the mean square of the calculated regression line (S1012-4). The computer then calculates the slope of the regression line (S1012-5), obtains the calculated slope as α, and terminates the process.

[0052] The computer evaluates α as a feature of the fluctuation waveform. For example, if α is approximated to 0.5, it indicates white noise; if it is approximated to 1, it indicates 1 / f fluctuation; and if it is approximated to 1.5, it indicates Brownian noise. When α is approximated to 1, the computer considers the time-series data of chewing intervals to exhibit 1 / f fluctuation and determines that the subject was performing unconscious chewing movements.

[0053] <3. Regarding the judgment process S102> The determination process S102 will now be explained. The case where fluctuation analysis is performed in each of the two fluctuation analysis processes S101 described above will be explained. In the determination process S102, the computer determines the chewing state of the subject person. The chewing state indicates, for example, the subject person's conscious state during chewing, indicating whether they are chewing consciously or unconsciously. In other words, the determination process S102 determines whether the result of the fluctuation analysis process S101 (the slope of the regression line) shows 1 / f fluctuation, and determines the chewing state of the subject person according to the determination result. In the following explanation, the determination process S102 determines whether the chewing state is conscious or unconscious, but for example, it may also determine whether it is conscious or not, or unconscious or not.

[0054] <3.1 When fluctuation analysis is performed using the fluctuation analysis process S1011 based on power spectral density> The computer determines that the time-series data of chewing intervals exhibits 1 / f fluctuation when β approximates -1. For example, the computer determines that 1 / f fluctuation is present when β is less than -0.9 and greater than -1.1 (-0.9 > β > -1.1).

[0055] Figure 9 shows examples of graphs of β per unit time during single-task and dual-task execution. Single-task execution is, for example, when the subject performs only the chewing action. In contrast, dual-task execution is when the subject performs one or more actions in parallel with the chewing action.

[0056] Figure 9 shows the values ​​of β when the task is a single task, consisting only of chewing gum, and when the task is a dual task, consisting of chewing gum in addition to solving a calculation problem. According to Figure 9, the value of β in the single task is between -1.2 and -1.3. The computer determines that the β of the time-series data of the chewing interval in the single task is outside the predetermined range (-0.9 > β > -1.1), and therefore does not show 1 / f fluctuation (No. in judgment process S102 in Figure 1), and determines that it is a conscious chewing movement (process S103 in Figure 1).

[0057] On the other hand, as shown in Figure 9, β in the dual-task case shows a value around -1.0. The computer determines that the time-series data of chewing intervals in the dual-task case shows 1 / f fluctuation (Yes in judgment process S102 in Figure 1) because β is within a predetermined range (-0.9 > β > -1.1), and therefore determines that it is an unconscious chewing movement (process S104 in Figure 1).

[0058] Thus, by acquiring chewing data for single-task and dual-task subjects, performing trend-removed fluctuation analysis, and comparing the results (evaluation process), it can be seen that the chewing data for dual-task subjects exhibits a gradient closer to 1 / f fluctuation. In other words, the chewing data for single-task subjects does not show 1 / f fluctuation, indicating conscious chewing movements, while the chewing data for dual-task subjects shows 1 / f fluctuation, indicating unconscious chewing movements.

[0059] <3.2 When fluctuation analysis is performed using the trend removal fluctuation analysis process S1012> The computer determines that the time-series data of chewing intervals exhibits 1 / f fluctuation when α is approximately equal to 1. For example, the computer determines that 1 / f fluctuation is present when α is greater than 0.95 and less than 1.05 (1.05 > α > 0.95).

[0060] Figure 10 shows examples of graphs of α per unit time during single-task and dual-task execution.

[0061] According to Figure 10, α in the single-task case shows a value of around 1.1. The computer determines that the α of the time-series data of chewing intervals in the single-task case is outside the predetermined range (1.05 > α > 0.95), and therefore does not show 1 / f fluctuation (No. in judgment process S102 in Figure 1), and determines that it is a conscious chewing movement (process S103 in Figure 1).

[0062] On the other hand, as shown in Figure 10, α in the dual-task case shows a value around 1.0. The computer determines that the α of the time-series data of chewing intervals in the dual-task case is within a predetermined range (1.05 > α > 0.95), thus exhibiting 1 / f fluctuation (Yes in judgment process S102 in Figure 1), and that it is an unconscious chewing movement (process S104 in Figure 1).

[0063] Thus, by acquiring chewing data for single-task and dual-task subjects, performing trend-removed fluctuation analysis, and comparing the results (evaluation process), it can be seen that the chewing data for dual-task subjects exhibits a gradient closer to 1 / f fluctuation. In other words, the chewing data for single-task subjects does not show 1 / f fluctuation, indicating conscious chewing movements, while the chewing data for dual-task subjects shows 1 / f fluctuation, indicating unconscious chewing movements.

[0064] In the first embodiment, evaluation of chewing data for single and dual tasks revealed that, in both cases, the chewing data for the single task did not exhibit 1 / f fluctuation, while the chewing data for the dual task did exhibit 1 / f fluctuation. Utilizing this characteristic, the computer can determine the state of chewing (whether it is conscious or unconscious) with high accuracy in the chewing movement determination process S10.

[0065] [Other embodiments] The chewing movement determination process may be used, for example, to evaluate food. For instance, studies have reported that conscious chewing in older adults can improve short-term memory and quality of life. It is expected that a person's conscious state during chewing will change depending on differences in various elements of food (hardness, taste, elasticity, size, etc.). By evaluating food using the chewing movement determination process in the first embodiment, researchers can develop foods and ingredients intended to encourage more conscious or more unconscious chewing. For example, this could involve developing foods that promote conscious chewing in older adults, or foods that promote unconscious chewing with a relaxing effect.

[0066] Specifically, the computer acquires time-series data of chewing intervals for multiple foods, performs fluctuation analysis on each time-series data, and extracts foods whose regression line slope does not show 1 / f fluctuation as foods that are consciously chewed (extraction step). The time-series data of chewing intervals is preferably, for example, data from a dual-task environment. [Explanation of Symbols]

[0067] S10: Chewing motion determination process S100: Time-series data acquisition process for chewing intervals S1001: Processing to acquire chewing intervals using image data S1002: Chewing interval acquisition process using surface electromyography of the masseter muscle. S101: Fluctuation analysis processing S1011: Fluctuation analysis processing using power spectral density S1012: Trend removal fluctuation analysis processing P1: Facial feature points P2: Facial feature points

Claims

1. The process involves acquiring time-series data of the chewing intervals of the subject person, The analysis process involves analyzing the waveform of fluctuations in the aforementioned time series data and calculating the slope of the regression line. A determination step is made to determine the chewing state of the subject person depending on whether the gradient exhibits 1 / f fluctuation or not. A method for determining chewing movements to be executed by a computer.

2. In the determination step, if the gradient does not exhibit 1 / f fluctuation, it is determined that the subject person is chewing consciously. The method for determining masticatory movement according to claim 1.

3. The acquisition process described above is: A shooting step to acquire image data of the subject person chewing for a predetermined period of time, A detection step of detecting the face of the target person from the image data and detecting multiple facial feature points from the detected face, The process includes a data acquisition step of determining the chewing state based on the amount of movement of the facial feature points per unit time and acquiring the time-series data. The method for determining masticatory movement according to claim 1.

4. The acquisition process described above is: A measurement step in which the activity level of the masseter muscle of the subject person is obtained by measuring surface electromyography during chewing of the subject person, The process includes a data acquisition step which determines the state of chewing based on the activity level and acquires the time-series data. The method for determining masticatory movement according to claim 1.

5. The determination step determines that if the gradient exhibits a 1 / f fluctuation, the subject person's chewing is performed unconsciously. The method for determining masticatory movement according to claims 1 to 2.

6. The process involves acquiring time-series data of the chewing intervals of a subject person while they chew food, and The analysis process involves analyzing the waveform of fluctuations in the aforementioned time series data and calculating the slope of the regression line. A determination step is made to determine whether the food is chewed consciously or unconsciously, depending on whether the gradient exhibits 1 / f fluctuation. A method for determining food quality using a computer.

7. The acquisition process involves obtaining first time-series data of the chewing interval when the subject chews while having no other issues besides chewing, and second time-series data of the chewing interval when the subject chews while having other issues besides chewing. The analysis process involves analyzing the fluctuation waveforms of the first time series data and the second time series data, and calculating the first slope of the regression line for the first time series data and the second slope of the regression line for the second time series data. An evaluation step is performed in which the first gradient and the second gradient are compared with the gradient in 1 / f fluctuation, and the conscious state during chewing is evaluated. A method for evaluating chewing movements performed by a computer.

Images