JP6048242B2 - Eating motion detection device, eating motion detection method and program - Google Patents

Description

  The present disclosure relates to a food movement detection device, a food movement detection method, and a program.

  2. Description of the Related Art Conventionally, a food movement detection system that determines the presence or absence of a user's food movement using a three-axis acceleration sensor attached to the lower arm of both arms of the user is known (see, for example, Patent Document 1).

JP 2011-115508 A

  However, the movement of the user's arm during the eating movement is similar to the movement of the user's arm other than the eating movement (for example, movement of both arms during the action of biting the nose, rubbing the eyes, taking an object, etc.). There are many cases. In addition, the movement of the arm differs depending on whether the food is in the vicinity, the food is in the distance, the food is cut immediately before, the noodles are rinsed and eaten, etc. There are many cases. In addition, various other operations such as picking food, cutting food with a knife or chopsticks, draining noodles, etc. may occur immediately before the eating operation. If the eclipse motion occurs continuously, it becomes difficult to detect the eclipse motion. Therefore, the triaxial acceleration sensor attached to the lower arm of both arms of the user as in the configuration described in Patent Document 1 has a problem that the eating operation cannot be detected with high accuracy.

  Therefore, an object of the disclosed technique is to provide a eating motion detection device, a eating motion detection method, and a program that can accurately detect a eating motion.

According to one aspect of the present disclosure, an acceleration sensor worn on the chest of the user or on the upper part of the abdomen,
A processing device that determines the presence or absence of a eating action, which is an action of the user carrying food to the mouth, based on an output signal of the acceleration sensor ;
The processing device determines that the user's eating action is present when a series of movements of the user such as “still, bent forward, still, back, rest” is detected based on an output signal of the acceleration sensor. An eating motion detection device is provided.

  According to the technique of the present disclosure, a eating action detection device, an eating action detection method, and a program that can detect eating action with high accuracy are obtained.

It is a figure which shows an example of schematic structure of the eating movement detection apparatus. It is a figure which shows an example of the definition of the axial direction of the acceleration sensor. It is a figure which shows an example of the definition of the positive / negative direction of the detection axis of the acceleration sensor. 2 is a diagram illustrating an example of a hardware configuration of a processing apparatus 100. FIG. It is a figure which shows an example of the acceleration data (time series of an output signal) of the acceleration sensor 10 obtained at the time of a eating operation. As a comparison, it is a figure which shows an example of the acceleration data obtained at the time of eating operation when the triaxial acceleration sensor with which a user's arm is mounted | worn is used. It is a figure which shows an example of the eating action pattern which may be used in order to detect the waveform pattern peculiar at the time of eating action. It is explanatory drawing of an example of the calculation method of a correlation coefficient. It is explanatory drawing of an example of the setting method of the predetermined threshold value Tz. It is a figure which shows an example of the relationship between the waveform pattern of the vertical acceleration GX obtained at the time of eating operation, and the waveform pattern of the longitudinal acceleration GZ obtained at the time of eating operation. It is a figure which shows an example of the relationship between the waveform pattern of the left-right acceleration GY obtained at the time of eating operation, and the waveform pattern of the longitudinal acceleration GZ obtained at the time of eating operation. 3 is a functional block diagram illustrating an example of functions of the processing apparatus 100. FIG. 4 is a flowchart illustrating an example of processing executed by the processing device 100. It is a figure which shows the example of a data division. It is a figure which shows the example of the result (the 1) obtained in the several steps of the process of FIG. It is a figure which shows the example of the result (the 2) obtained in the several steps of the process of FIG. It is explanatory drawing of the usage example of the process shown in FIG. It is a flowchart which shows an example of the eating operation | movement determination process (the 1) performed by the processing apparatus. It is a flowchart which shows an example of the continuation (the 2) of the eating action determination process shown in FIG. It is a figure which shows an example of the division | segmentation of the acceleration data of the longitudinal acceleration GZ, and an example of the time series data of the longitudinal acceleration after a division | segmentation. It is a figure which shows the example of data of a eating action pattern. It is a figure which shows the example of calculation of a correlation coefficient. It is a figure which shows an example of the acquisition range of the time series of the vertical acceleration GX.

  Hereinafter, embodiments will be described in detail with reference to the accompanying drawings.

  FIG. 1 is a diagram illustrating an example of a schematic configuration of the eating motion detection device 1.

  The eating motion detection device 1 includes a processing device 100 and an acceleration sensor 10.

  The processing apparatus 100 may be configured by an arbitrary form of computer. Various functions (including functions described below) of the processing device 100 may be realized by arbitrary hardware, software, firmware, or a combination thereof. For example, any or all of the functions of the processing apparatus 100 may be realized by an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA), or the like. Further, the processing device 100 may be realized by a plurality of processing devices. In the example illustrated in FIG. 1, the processing apparatus 100 is in the form of a server (host computer) located on the cloud 2.

  The acceleration sensor 10 has an arbitrary form that can be worn on the chest of the user. For example, the acceleration sensor 10 may be configured to be stored in a chest pocket of a user's clothing, or may be fixed to the user's chest with a dedicated belt or the like, or may be a clip or the like. It may be attached to the chest of the clothes. The acceleration sensor 10 can communicate with the processing device 100 via a wireless communication network. For example, the acceleration sensor 10 may perform wireless communication with the processing device 100 based on standards such as Bluetooth (registered trademark) and WiMAX (Worldwide Interoperability for Microwave Access).

  The acceleration sensor 10 acquires acceleration data in the user's chest. The acceleration sensor 10 transmits the acquired acceleration data to the processing device 100. As shown in FIG. 1, the acceleration sensor 10 may directly transmit acceleration data to the processing device 100 or may transmit the acceleration data to the processing device 100 via another communication terminal 30 such as a smartphone.

  Note that the processing apparatus 100 may be realized by a computer 20 (or a computer in a mobile terminal) instead of the server located on the cloud 2. In this case, the acceleration sensor 10 may transmit acceleration data to the computer 20 via wireless communication or wired communication. Further, the processing apparatus 100 may be mounted in the acceleration sensor 10. Also, some or all of the functions of the processing apparatus 100 may be realized in cooperation with a plurality of physically separate processing apparatuses. For example, some or all of the functions of the processing device 100 may be realized in cooperation with the processing device in the acceleration sensor 10 and the computer 20 (or a server located on the cloud 2).

  FIG. 2 is a diagram illustrating an example of the definition of the axial direction of the acceleration sensor 10. Here, as an example, as shown in FIG. 2, the X axis corresponds to the vertical direction (vertical direction) of the user, the Y axis corresponds to the horizontal direction (horizontal direction) of the user, and the Z axis corresponds to the user. Corresponds to the front-rear direction (front / rear direction). The X axis, the Y axis, and the Z axis are orthogonal to each other. The acceleration sensor 10 detects these triaxial accelerations, that is, triaxial accelerations orthogonal to each other. Here, it is assumed that the three axes of the acceleration sensor 10 are in a fixed relationship with the user's chest. However, when the acceleration sensor 10 is stored in clothes or the like, the complete fixed relationship may be lost due to the movement of the clothes with respect to the chest of the user, but such a point is not considered (can be ignored).

  FIG. 3 is a diagram illustrating an example of the definition of the positive / negative direction of the detection axis of the acceleration sensor 10. 3A is a diagram schematically showing a user wearing the acceleration sensor 10 in a top view along the X axis, and FIG. 3B is a diagram schematically showing the user in a side view along the Y axis. FIG.

  In the following, as shown in FIG. 3, the Z-axis will be described with the user's rear side as “positive”, and the X-axis will be described with the user's upper side as “positive”. In this case, for example, the acceleration sensor 10 may be housed in the housing so that the normal direction outward direction of one surface of the housing of the acceleration sensor 10 is the positive direction of the Z axis. In this case, the user should just wear | wear so that the said one surface side of a housing | casing may face a chest. In addition, the definition of the positive / negative direction shown in FIG. 3 is an example to the last, and the reverse setting is also possible.

  FIG. 4 is a diagram illustrating an example of a hardware configuration of the processing apparatus 100.

  In the example illustrated in FIG. 4, the processing device 100 includes a control unit 101, a main storage unit 102, an auxiliary storage unit 103, a drive device 104, a network I / F unit 106, and an input unit 107.

  The control unit 101 is an arithmetic device that executes a program stored in the main storage unit 102 or the auxiliary storage unit 103, receives data from the input unit 107 or the storage device, calculates, processes, and outputs the data to the storage device or the like. To do.

  The main storage unit 102 is a ROM (Read Only Memory), a RAM (Random Access Memory), or the like, and a storage device that stores or temporarily stores programs and data such as an OS and application software that are basic software executed by the control unit 101. It is.

  The auxiliary storage unit 103 is an HDD (Hard Disk Drive) or the like, and is a storage device that stores data related to application software or the like.

  The drive device 104 reads the program from the recording medium 105, for example, a flexible disk, and installs it in the storage device.

  The recording medium 105 stores a predetermined program. The program stored in the recording medium 105 is installed in the processing device 100 via the drive device 104. The installed predetermined program can be executed by the processing apparatus 100.

  The network I / F unit 106 includes a peripheral device (for example, the acceleration sensor 10 or the like) having a communication function connected via a network constructed by a data transmission path such as a wired and / or wireless line, and the processing device 100. Interface.

  The input unit 107 includes a keyboard having cursor keys, numeric input, various function keys, and the like, a mouse, a slice pad, and the like.

  In the example illustrated in FIG. 4, various processes described below can be realized by causing the processing apparatus 100 to execute a program. It is also possible to record the program on the recording medium 105 and cause the processing apparatus 100 to read the recording medium 105 on which the program is recorded, thereby realizing various processes described below. The recording medium 105 is a recording medium that records information optically, electrically, or magnetically, such as a CD-ROM, a flexible disk, or a magneto-optical disk, and information is electrically stored such as a ROM or flash memory. Various types of recording media such as a semiconductor memory for recording can be used. Note that the recording medium 105 does not include a carrier wave.

  FIG. 5 is a diagram illustrating an example of acceleration data (time series of output signals) of the acceleration sensor 10 obtained during the eating operation. FIG. 6 is a diagram showing an example of acceleration data obtained during a eating operation when a three-axis acceleration sensor worn on the user's arm is used as a comparison.

  In FIGS. 5 and 6, waveform portions when the eating operation is performed are indicated by A 1, A 2, A 3 and A 1 ′, A 2 ′, A 3 ′. A1 and A1 ′ are related to the normal eating movement that takes food to the mouth after grabbing food, A2 and A2 ′ are related to an irregular eating action that carries food to the mouth while turning chopsticks, and A3 and A3 ′ are near This is related to the eating action that is carried to the mouth after grabbing the food in

  First, as a comparative example, in the case of a triaxial acceleration sensor worn on an arm, a characteristic waveform is generated during an eating operation as shown in FIG. However, as shown in A1 ′, A2 ′, A3 ′, the movement of the arm (characteristic waveform) varies according to various eating actions. Furthermore, although illustration is omitted, the movement of the user's arm during the eating operation is the movement of the user's arm during the operation other than the eating operation (for example, when acting such as biting the nose, rubbing eyes, taking an object, etc.) The movement of both arms) is often similar. Therefore, it can be understood that it is difficult to accurately detect the eating action when using a three-axis acceleration sensor attached to the arm.

  On the other hand, in the case of the acceleration sensor 10 attached to the chest, a characteristic waveform is generated during the eating operation as shown in FIG. For example, with respect to the longitudinal acceleration GZ, as shown in A1, A2, and A3, a characteristic waveform pattern appears when performing the eating operation. Specifically, with respect to the longitudinal acceleration GZ, a common waveform pattern of “constant value, decrease, constant value, increase, constant value” is seen in A1, A2, and A3. This corresponds to a series of operations such as “still, bend forward, still, return, rest” when the user performs an eating operation.

  Therefore, as can be seen from FIG. 5, the Z-axis component of the acceleration signal from the acceleration sensor 10 during the eating operation corresponds to a series of operations of “stationary, bent forward, stationary, back, stationary”, respectively. A characteristic waveform is generated. Since this series of movements is chest movement, it occurs in common with various eating movements as opposed to arm movement. Therefore, the processing apparatus 100 can detect the eating action with high accuracy by detecting this characteristic waveform pattern.

  In the case of the acceleration sensor 10 attached to the chest, as shown in FIG. 5, the same tendency as the characteristic waveform of the longitudinal acceleration GZ appears in the vertical acceleration GX. Specifically, the vertical acceleration GX decreases in synchronization with the decrease in the longitudinal acceleration GZ, and the vertical acceleration GX increases in synchronization with the increase in the longitudinal acceleration GZ after the decrease. This is due to a change in the inclination (angle with respect to the vertical direction) of the chest of the user when “bend forward” and “return back”. Such a feature also occurs in common with various eating movements as opposed to arm movements because of the movement of the chest. Therefore, the processing apparatus 100 can detect the eating operation with high accuracy by further detecting this characteristic waveform pattern.

  Further, in the case of the acceleration sensor 10 attached to the chest, as shown in FIG. 5, a characteristic waveform pattern appears when performing the eating operation that there is substantially no change in the lateral acceleration GY. This is because the eating operation does not involve the operation of tilting the chest left and right. In contrast, in the case of a triaxial acceleration sensor attached to the arm, as shown in FIG. 6, various change patterns are generated in the acceleration waveform of each axis. Therefore, the processing apparatus 100 can detect the eating operation with high accuracy by further detecting this characteristic waveform pattern.

  The waveform pattern peculiar to the eating operation as described above may be detected by an arbitrary method. For example, it may be detected using a plurality of threshold values in time series, or may be detected by pattern matching.

  Next, an example of a method for detecting the waveform pattern of the longitudinal acceleration GZ that is characteristic during the eating operation as described above will be described. In this example, a predetermined waveform pattern (eating motion pattern) corresponding to the waveform pattern of the longitudinal acceleration GZ peculiar to the eclipse motion is created in advance, and the correlation coefficient between the eclipse motion pattern and the actually obtained acceleration signal is calculated. A method for detecting the eating motion by calculating will be mainly described.

  FIG. 7 is a diagram illustrating an example of the eating action pattern.

  The correlation coefficient with the time series of the longitudinal acceleration actually obtained during the eating motion is significantly higher than the correlation coefficient with the time series of the longitudinal acceleration actually obtained during other times than the eating motion. Is generated. In the example illustrated in FIG. 7, the eating motion pattern is “constant, lowering, constant, rising, constant” corresponding to a series of motions during the eating operation of “still, bending forward, resting, returning, resting”. Includes patterns.

In FIG. 7, Ta, Tb, Tc, L ₁ , and L ₂ are parameters that determine the eating operation pattern, Ta, Tb, and Tc are times, and L ₁ and L ₂ are amplitudes (pattern signal values). is there. Specifically, Ta represents the first “constant” time width of “constant, lowering, constant, rising, constant”, Tb represents the second “constant” time width, and Tc Represents the third “constant” time span. L ₁ and L ₂ represent the amplitude reduction range at the time of “decrease” of “constant, decrease, constant, increase, constant”, and L ₂ represents “constant, decrease, constant, increase, constant”. Of these, the amplitude rises when “rise”.

Parameters Ta, Tb, Tc, L ₁ , and L ₂ are the values when the correlation coefficient between the time series of the longitudinal acceleration actually obtained during the eating operation and the eating operation pattern is the longitudinal acceleration actually obtained during other than the eating operation. It is set so as to be significantly higher than the correlation coefficient between the series and the eating action pattern. For example, the parameters Ta and Tc may be arbitrary values within the range of 1 second to 4 seconds. The parameter Tb may be an arbitrary value within the range of 0.4 seconds to 20 seconds. The parameters L ₁ and L ₂ may be any value within the range of 1500 mG to 8000 mG.

A plurality of eating motion patterns are preferably generated. This is because the details (for example, parameters Ta, Tb, Tc, L ₁ , and L ₂ ) may be different even though they are the same as a series of operations of “still, bend forward, still, return, and still”. It is. For this reason, a plurality of eating motion patterns may be generated based on acceleration signals actually obtained during various eating operations. At this time, the eating action pattern may be generated for each user in consideration of individual differences, or may be generated based on data from a plurality of users.

  FIG. 8 is an explanatory diagram of an example of a correlation coefficient calculation method. In FIG. 8, D1 shows an example of a time series of longitudinal acceleration in a certain period. D1 may correspond to a part of a time series of longitudinal acceleration obtained by moving a window (window) having a constant window width W1. D2 shows an example of the eating action pattern. Hereinafter, the acceleration data in the window having the constant window width W1 is also referred to as “window data”.

  The correlation coefficient may be calculated between the window data and the eating action pattern. At this time, since the time width of the window data and the eating operation pattern may be different, even if a plurality of correlation coefficients are calculated while shifting the shorter one of the window data and the eating operation pattern by a predetermined time. Good. In the example shown in FIG. 8, since the window data is longer than the time width of the eating operation pattern, the eating operation pattern is shifted to the right by a predetermined time, as shown by P1 and P2 in FIG. A correlation coefficient is calculated. As a result, as shown in FIG. 8, a plurality of correlation coefficients (corresponding to the number shifted) are calculated for one window data.

  The correlation coefficient has a high value when the similarity (correlation) between the window data and the eating action pattern is large. For example, the correlation coefficient cor may be calculated based on the following equation.

ここで、Ｘは、窓データの信号を表し、Ｙは、食動作パターンの信号を表す。また、mean（Ｘｉ）は、データＸの相加平均であり、mean（Ｙｉ）は、データＹの相加平均である。

Here, X represents a window data signal, and Y represents a food movement pattern signal. Mean (Xi) is an arithmetic mean of data X, and mean (Yi) is an arithmetic mean of data Y.

  In this case, when the correlation coefficient is equal to or greater than the predetermined threshold Tz, it may be determined that the shape is similar (the waveform corresponding to the eating action pattern exists). When a plurality of correlation coefficients are obtained for one window data, the maximum value of the correlation coefficient is defined as the shape similarity, and when the shape similarity is equal to or greater than a predetermined threshold Tz, it is determined that the shapes are similar. It's okay. In this case, the data range (P1, P2, etc.) in the window data when the correlation coefficient becomes the maximum value is also referred to as “a window period determined to be similar to the pattern”.

  In addition, although the fixed window width W1 of the window which determines the time width of window data is arbitrary, for example, it may be the same as the maximum time width of a plurality of eating motion patterns. Alternatively, the constant window width W1 may be significantly longer than the maximum time width among the plurality of eating motion patterns. When the constant window width W1 is long, the maximum value of the correlation coefficient (greater than or equal to the predetermined threshold Tz) can occur at a plurality of positions separated by the time width of the eating action pattern in the window data. This means that the eating operation can exist a plurality of times in the window data. In this case, there are a plurality of window periods determined to be pattern similar.

  FIG. 9 is an explanatory diagram of an example of a method for setting the predetermined threshold Tz.

  The predetermined threshold value Tz is a threshold value for partitioning the correlation coefficient at the time of eating operation and the correlation coefficient at the time of other operation (including no operation time), and may be set in an arbitrary manner. For example, the predetermined threshold Tz may be set empirically or may be adapted based on a learning result (test result).

When based on the learning result, for example, the predetermined threshold Tz may be derived as follows. First, a plurality of time series of longitudinal accelerations actually obtained at the time of eating operation and a plurality of time series of longitudinal accelerations actually obtained at times other than the eating operation (during other operations) are prepared. The time width of the time series data is arbitrary, but may be a time width corresponding to the time width of the window data, for example. Next, with respect to the time series of longitudinal acceleration actually obtained during the eating operation, the shape similarity with all the eating operation patterns is calculated, and the maximum value SimMax is obtained. This is executed for each of the prepared time series of longitudinal accelerations during the eating operation. Similarly, with respect to the time series of longitudinal accelerations actually obtained during other motions, the shape similarity with all eating motion patterns is calculated, and the maximum value SimMax is obtained. This is executed for each of a plurality of prepared time series of longitudinal acceleration during other operations. Then, to calculate the average _{m M} and standard deviation sigma _M of SimMax during food operation, it calculates the average _{m O} and standard deviation sigma _O of SimMax during other operations. Then, the point at which the Mahalanobis distance between the eating operation and the other operation becomes equal is obtained as the threshold value Tz. That is, a threshold value Tz that satisfies the following expression is calculated.

尚、同様に、閾値Ｔｚは、個人差を考慮してユーザ毎に生成されてもよいし、複数のユーザからのデータに基づいて生成されてもよい。

Similarly, the threshold value Tz may be generated for each user in consideration of individual differences, or may be generated based on data from a plurality of users.

  Further, the waveform pattern of the longitudinal acceleration GZ that is peculiar during the eating operation has a predetermined amplitude characteristic as shown in FIG. The amplitude characteristic may simply be the maximum amplitude, minimum amplitude, average amplitude, etc. of the longitudinal acceleration GZ, or the difference between the maximum value of the longitudinal acceleration GZ and the minimum value of the longitudinal acceleration GZ (hereinafter referred to as “maximum / minimum difference”). May be used). Therefore, the waveform pattern of the longitudinal acceleration GZ peculiar to the eating operation may be detected using such an amplitude feature. For example, when the time-series maximum / minimum difference of the longitudinal acceleration GZ is within a predetermined range, it may be determined that the amplitude condition in the longitudinal direction is satisfied. The predetermined range may be set by an arbitrary method, but may be set by the following method, for example. First, a plurality of time series of longitudinal acceleration GZ actually obtained at the time of eating operation are prepared. Next, a maximum / minimum difference is calculated for each time series data, and is set based on the minimum value and the maximum value. That is, the predetermined range may be set so as to correspond to a range in which a time-series maximum / minimum difference of the longitudinal acceleration GZ actually obtained during the eating operation can be taken. Note that the predetermined range may be determined for each user in consideration of individual differences, or may be determined based on data from a plurality of users.

  Next, an example of a method for detecting a waveform pattern of the vertical acceleration GX that is peculiar during the eating operation will be described.

  FIG. 10 is a diagram illustrating an example of the relationship between the waveform pattern of the vertical acceleration GX obtained during the eating operation and the waveform pattern of the longitudinal acceleration GZ obtained during the eating operation.

  As described above, the waveform pattern of the vertical acceleration GX obtained during the eating operation has a certain correlation with the waveform pattern of the longitudinal acceleration GZ obtained during the eating operation, as shown in FIG. For example, the “bend forward” operation during the eating operation occurs “forward” and “downward”, so that the vertical acceleration GX and the longitudinal acceleration GZ both change in the negative direction. That is, there is a positive correlation between the vertical acceleration GX and the longitudinal acceleration GZ.

  Whether there is a positive correlation between the vertical acceleration GX and the longitudinal acceleration GZ may be determined by monitoring the change directions of both the vertical acceleration time series and the longitudinal acceleration time series. The determination may be made by calculating the correlation coefficient between the series and the time series of the longitudinal acceleration. When calculating the correlation coefficient, the correlation coefficient may be calculated between the entire window data including the window period determined to be similar to the pattern. Alternatively, the correlation coefficient may be calculated between data in window periods determined to be pattern similar.

  Further, the waveform pattern of the vertical acceleration GX that is peculiar to the eating operation has a predetermined amplitude characteristic as shown in FIG. The amplitude feature may simply be the maximum amplitude, minimum amplitude, average amplitude, etc. of the vertical acceleration GX, or may be the difference (maximum minimum difference) between the maximum value of the vertical acceleration GX and the minimum value of the vertical acceleration GX. Good. Therefore, the waveform pattern of the vertical acceleration GX peculiar to the eating operation may be detected using such an amplitude feature. For example, when the time-series maximum / minimum difference of the vertical acceleration GX is within a predetermined range, it may be determined that the vertical amplitude condition is satisfied. The predetermined range may be set by an arbitrary method, but may be set by the following method, for example. First, a plurality of time series of vertical accelerations actually obtained at the time of eating operation are prepared. Next, a maximum / minimum difference is calculated for each time series, and is set based on the minimum value and the maximum value. In other words, the predetermined range may be set so as to correspond to a range in which the maximum and minimum differences in the time series of vertical acceleration actually obtained during the eating operation can be taken. Note that the predetermined range may be determined for each user in consideration of individual differences, or may be determined based on data from a plurality of users.

  Next, an example of a method for detecting a waveform pattern of the left-right acceleration GY that is peculiar during eating is described.

  FIG. 11 is a diagram illustrating an example of the relationship between the waveform pattern of the lateral acceleration GY obtained during the eating operation and the waveform pattern of the longitudinal acceleration GZ obtained during the eating operation. As described above, the waveform pattern of the lateral acceleration GY obtained during the eating operation hardly changes during the eating operation as shown in FIG. That is, the variance value of the lateral acceleration GY during the eating operation is low. Therefore, if the variance value of the lateral acceleration GY does not exceed the predetermined threshold value Ty, it may be determined that the lateral eating condition is satisfied. The variance value of the left-right acceleration GY used for this determination may be calculated within the entire window data including the window period determined to be pattern similar, or may be calculated within the window period determined to be pattern similar. Good.

  The predetermined threshold value Ty may be set by any method, but may be set by the following method, for example. First, a plurality of time series of left and right accelerations actually obtained at the time of eating operation are prepared. Next, a variance value is calculated for each time series, and the maximum value is set as the threshold value Ty. That is, the predetermined threshold value Ty may be set so as to correspond to the maximum value of a range that can be taken by the variance value of the lateral acceleration actually obtained during the eating operation. Note that the predetermined threshold value Ty may be determined for each user in consideration of individual differences, or may be determined based on data from a plurality of users.

  Next, a specific example of the function (processing) of the processing apparatus 100 will be described.

  FIG. 12 is a functional block diagram illustrating an example of functions of the processing apparatus 100.

  In the example illustrated in FIG. 12, the processing apparatus 100 includes a Z-axis acceleration data acquisition unit 152, a data division unit 154, an amplitude condition determination unit 156, a shape determination unit 158, and an X / Y-axis acceleration data acquisition unit 160. The vertical direction operation condition determination unit 162 and the horizontal direction operation condition determination unit 164 are included. In addition, the processing apparatus 100 includes a eating action frequency calculation unit 166, a meal period candidate detection unit 168, a temporal validity determination unit 170, and a result holding unit 172. In addition, the processing apparatus 100 includes an observation acceleration data holding unit 180, a eating action pattern data holding unit 182, and a threshold data holding unit 184. The observation acceleration data holding unit 180, the eating movement pattern data holding unit 182 and the threshold data holding unit 184 may be realized by an arbitrary storage device.

  FIG. 13 is a flowchart illustrating an example of processing executed by the processing device 100. FIG. 14 is an explanatory diagram of FIG. 13 and illustrates an example of data division. FIGS. 15 and 16 are explanatory diagrams of FIG. 13 and show examples of results obtained at several stages of the processing of FIG. The process shown in FIG. 13 is executed every other day as an example, but may be executed with a shorter cycle or may be executed with a longer cycle.

  In step 1300, the Z-axis acceleration data acquisition unit 152 acquires acceleration data for one day accumulated in the observation acceleration data holding unit 180. Note that acceleration data obtained from the acceleration sensor 10 is stored in the observation acceleration data holding unit 180. The observation acceleration data holding unit 180 may accumulate acceleration data for several days. The output signal from the acceleration sensor 10 may be stored in the observation acceleration data holding unit 180 after being subjected to a predetermined process (for example, a filter process).

  In step 1302, the data dividing unit 154 divides the acceleration data for one day acquired by the Z-axis acceleration data acquiring unit 152 by the constant window width W1 by moving the window (window) having the constant window width W1. . Thereby, a plurality of window data having a certain window width W1 is obtained. FIG. 14 shows the i-th window data and the i + 1-th window data. In the example shown in FIG. 14, the window is shifted every fixed window width W1, but may be shifted every width smaller than the fixed window width W1 (for example, about half of the fixed window width W1).

  In step 1304, i = 1 is set. Note that i is a window data number.

  In step 1306, it is determined whether or not there is an eating operation for the i-th window data. Whether or not there is an eating operation may be determined according to whether or not a specific waveform pattern is detected during the eating operation as described above. A specific example of the processing in step 1306 will be described later.

  In step 1308, it is determined whether i is the last. If i is the last, the process proceeds to step 1310. Otherwise, i = i + 1 is updated, and the process returns to step 1304. In this way, it is determined whether or not there is an eating operation for all the window data. As a result, a determination result as shown in FIG. The example shown in FIG. 15A is a case where there are N pieces of window data, No represents that it is determined that there is no eating operation, and Yes indicates that it is determined that there is an eating operation. Represent.

  In step 1310, the eating action frequency calculation unit 166 calculates the eating action frequency for each period of the predetermined section width W2 based on the processing result in step 1308. In the example shown in FIG. 15B, as an example, the constant section width W2 is 10 minutes, the constant window width W1 is 40 seconds, and the window shift width is 20 seconds. In this case, the number of window data corresponding to 10 minutes is 29 (= {(10 × 60) / (40/2)} − 1). In the example shown in FIG. 15B, for the period number k, for example, the number of times that the eating operation is determined to be 15 is 15, so the eating operation frequency is 0.49 (= 15/29). Yes. In addition, for the period number K, the number of times that the eating operation is determined to be 0 is 0, and the eating operation frequency is 0 (= 0/29).

  In step 1312, the meal period candidate detection unit 168 determines whether or not each period for which the eating action frequency is calculated is a eating period based on the eating action frequency calculated in step 1310. For example, when the eating operation frequency is equal to or higher than the predetermined threshold frequency Ra, the eating period candidate detection unit 168 may determine the period related to the eating operation frequency as the eating period (candidate). This is because eating action tends to occur frequently during meals. The predetermined threshold frequency Ra may be set empirically or may be adapted based on a test result or the like. The predetermined threshold frequency Ra may be determined for each user in consideration of individual differences, or may be determined based on data from a plurality of users. As a result, a determination result as shown in FIG. The example shown in FIG. 15C is a case where there are M periods in which the eating motion frequency is calculated, No indicates that it is not a meal period (candidate), and Yes indicates a meal period. This indicates that it is determined to be a (candidate).

  In step 1314, when it is determined that the meal period candidate detection unit 168 is the meal period (candidate), the start time and the end time of the meal period (candidate) are acquired. In addition, when it determines with it not being a meal period, a start time and an end time do not need to be acquired. As a result, information on the start time and end time as shown in FIG.

  In step 1316, j = 1 is set. J is the number of the meal period (candidate).

  In step 1318, the temporal validity determination unit 170 determines the validity of the time as the meal period for the meal period (candidate) associated with the number j. This is based on the general tendency that a meal is basically performed within a predetermined time range and ends within a predetermined time range. For example, breakfast is within the time range from 6:00 to 8:00, lunch is within the time range from 11:00 to 14:00, dinner is within the time range from 18:00 to 24:00, and the meal time is 8 minutes to 30 minutes For example, minutes. In addition, these time ranges and meal time lengths are often accompanied by individual differences depending on business styles, home circumstances, and the like. Therefore, a questionnaire may be conducted in advance for the time zone and the meal length of each meal (breakfast, lunch, dinner), and the validity of the time may be determined based on the questionnaire results.

For example, the preliminary questionnaire may include the following four items.
-What time (T1a) to what time (T1b) to have breakfast?-What time (T2a) to what time (T2b) to have lunch?-What time (T3a) to what time (T3b) to have dinner? T3b) ・ How many minutes (Tp) or more and less than (Tq) the meal time length In this case, if the start time of the meal period (candidate) for number j is T0a and the end time is T0b , T1a <T0a, T0b <T1b, and Tp <(T0b-T0a) <Tq, it may be determined that “breakfast” is appropriate. If T2a <T0a, T0b <T2b, and Tp <(T0b-T0a) <Tq, it may be determined that “lunch” is appropriate. Further, if T3a <T0a, T0b <T3b, and Tp <(T0b-T0a) <Tq, it may be determined that “dinner” is valid.

  In step 1320, it is determined whether j is the last. If j is the last, the process proceeds to step 1322; otherwise, j = j + 1 is updated and the process returns to step 1316. In this way, the validity of time is determined for all meal periods (candidates).

  In step 1322, the determination result in step 1318 is output. As a result, a determination result as shown in FIG. 16B is obtained. The determination result is held in the result holding unit 172. In the example shown in FIG. 16B, there are four meal periods (candidates), candidate number 1 is determined to be breakfast, candidate number 2 is determined to be lunch, and candidate number 4 is determined to be dinner. It represents what has been done.

  In the process shown in FIG. 13, snacks and midnight snacks are not treated as meals, but they may be treated as meals. In the process shown in FIG. 13, all meals of breakfast, lunch, and dinner are detected, but only any one or only two may be detected.

  FIG. 17 is an explanatory diagram of a usage example of the process illustrated in FIG. 13, and is a diagram illustrating a display example of the processing result illustrated in FIG. 13.

  FIG. 17 shows a display representing a eating habit (eating habit) based on the processing result shown in FIG. Specifically, the history of each meal period of breakfast, lunch, and dinner is displayed for one week. Such a display may be displayed on a terminal carried by the user (for example, a mobile phone or a smartphone), a fixed terminal, or the like. For example, the processing shown in FIG. 13 may be realized by an application in a terminal carried by the user, and the processing result may be displayed on the display of the terminal as shown in FIG. In the example illustrated in FIG. 17, advice is output for the eating habit based on the processing result illustrated in FIG. 13. Such advice may be generated based on the processing result shown in FIG. Thereby, the user can confirm his / her eating habits afterwards and can obtain a direction to review his / her own eating habits.

By the way, in order to prevent lifestyle-related diseases such as metabolic syndrome and diabetes, we record daily lifestyle habits such as “exercise”, “meal”, “sleep”, etc., and will be aware of and improve their own problems. The process is important. In particular, “meal” preventive measures correspond to controlling “when”, “what” and “how much” as follows: Eat three meals regularly (when)
2. Have breakfast (when)
3. Nutritionally well balanced ("What")
4). Do not consume too many calories (how much)
5. Refrain from salt ("what")
Here, for example, if there is a long-term record of “when”, it is possible to detect irregular eating habits (for example, late dinner, no breakfast), and provide services such as providing preventive advice (FIG. 17). reference).

  In this respect, it is very troublesome to manually record the time of each meal. However, according to the process shown in FIG. 13, the meal period can be automatically recorded on a daily basis. Therefore, according to the process shown in FIG. 13 described above, it is possible to detect irregular eating habits (for example, late dinner, no breakfast), and provide appropriate preventive advice.

  Next, a specific example relating to the eating action determination process in step 1306 in the process shown in FIG. 13 will be described.

  FIG. 18 is a flowchart illustrating an example of the eating action determination process (part 1) executed by the processing apparatus 100. FIG. 19 is a flowchart showing an example of the continuation (part 2) of the eating action determination process shown in FIG. 20 to 23 are diagrams showing examples of results obtained at several stages of the processing shown in FIGS. 18 and 19, examples of data used in the processing shown in FIGS. 18 and 19, and the like.

  In step 1800, the i-th window data (time series of longitudinal acceleration GZ) is input. The window data is obtained by the data dividing unit 154 as described above (see step 1302 in FIG. 13). That is, the data dividing unit 154 divides the acceleration data of the longitudinal acceleration GZ by the constant window width W1 by moving the window (window) having the constant window width W1. In the example shown in FIG. 20A, m-th window data and m + 1-th window data are shown. In the example shown in FIG. 20A, unlike the example shown in FIG. 14, the window is shifted by a predetermined shift width smaller than the fixed window width W1 (for example, about half of the fixed window width W1). Yes. FIG. 20B shows M window data obtained in this way. In the example shown in FIG. 20B, the constant window width W1 corresponds to 40 seconds, and the sampling frequency is 200 ms / time.

  In Step 1802 and Step 1804, the amplitude condition determination unit 156 determines whether or not the time-series amplitude characteristics of the input longitudinal acceleration GZ are similar to the time-series amplitude characteristics of the longitudinal acceleration actually obtained during the eating operation. To do. Here, the difference amp between the maximum value and the minimum value of the longitudinal acceleration GZ (hereinafter also referred to as “maximum / minimum difference amp”) is used as an example of the amplitude feature.

  Specifically, in step 1802, the amplitude condition determination unit 156 subtracts the minimum value of the longitudinal acceleration GZ in the i-th window data from the maximum value of the longitudinal acceleration GZ in the i-th window data, thereby calculating the longitudinal acceleration GZ. Calculate the maximum and minimum difference amp.

  In step 1804, the amplitude condition determining unit 156 determines whether or not the maximum / minimum difference amp of the longitudinal acceleration GZ calculated in step 1802 is within a range of 1500 mG to 8000 mG. The range of 1500 mG to 8000 mG may correspond to a range that can be taken by the time series maximum / minimum difference amp of the longitudinal acceleration actually obtained during the eating operation, and may be adapted by a test or the like. This numerical range is merely an example, and may be modified as appropriate. If the maximum / minimum difference amp is in the range of 1500 mG to 8000 mG, it is determined that the amplitude condition is satisfied, and the process proceeds to step 1806. Otherwise, the process proceeds to step 1805.

  In step 1805, the amplitude condition determination unit 156 determines that there is no edible action in the window period related to the i-th window data because the amplitude condition is not satisfied. Thereby, window data related to other operations than the eating operation can be excluded. For example, an operation in which the maximum / minimum difference amp is significantly larger than that in the eating operation, such as an operation of picking up a fallen object, or an operation in which the maximum / minimum difference amp is significantly smaller than that in the eating operation, for example, an operation of pulling a chair. Such window data related to other operations can be eliminated.

  In step 1806, the shape determining unit 158 reads the value of the predetermined threshold Tz from the threshold data holding unit 184.

In step 1808, n = 1 is set. Note that n is the number of the eating action pattern. A plurality of eating motion patterns are prepared based on the five parameters Ta, Tb, Tc, L ₁ and L _{2 shown} in FIG. In the example shown in FIG. 21, N combinations of these five parameters (that is, N eating motion patterns) are prepared. In FIG. 21B, “NULL” means that there is no data and that the time width of the eating action pattern is short. For example, the eating action pattern related to the pattern N has data up to the time step T, whereas the eating action pattern in the pattern 1 and the pattern 2 has the data finished before the time step T (to the pattern N). The time width is shorter than the eating action pattern). The data table shown in FIG. 21B is held in the eating action pattern data holding unit 182.

  In step 1810, the shape determining unit 158 reads the eating action pattern Y (n) associated with the number n from the eating action pattern data holding unit 182. That is, the eating action pattern Y (n) corresponding to the number n is read from the data table shown in FIG.

  In step 1812, the shape determination unit 158 determines the time-series data length of the longitudinal acceleration GZ input in step 1800 (that is, the constant window width W1) and the data length of the eating motion pattern Y (n) read in step 1810. The smaller one is set as signal A and the larger one as signal B.

  In step 1814, k = 1 is set. Note that k indicates the start position of data used for calculating the correlation coefficient (see the description of step 1816).

In step 1816, the shape determination unit 158 acquires (extracts) a signal B ′ having the same data length as the signal A. This is to obtain a signal having the same data length necessary for calculating the correlation coefficient. The signal B ′ is as follows.
B ′ = B [k: k + L]
However, L corresponds to the data length of A. That is, the signal B ′ is data from the time step k to the time step k + L in the signal B.

  In step 1818, the shape determination unit 158 calculates a correlation coefficient cor (k) between the signal A and the signal B ′. The correlation coefficient cor (k) may be calculated based on the equation shown in the above equation 1. In this case, X corresponds to the signal A and Y corresponds to the signal B ′.

  In step 1820, shape determination unit 158 determines whether k + L matches the data length of signal B. If they do not match, k is changed to k + 1, and the processing of step 1816 and step 1818 is executed. In this way, the processes of Step 1816 and Step 1818 are repeated until k + L matches the data length of the signal B, and a plurality of correlation coefficients are acquired. FIG. 22 is a diagram illustrating an example of calculating a correlation coefficient, (A) illustrates an example of a time series of the longitudinal acceleration GZ, (B) illustrates an example of a eating motion pattern Y (n), C) shows an example of the correlation coefficient calculated from these. In the example shown in FIG. 22, the data length of the window data (time series of the longitudinal acceleration GZ) is 200 for the time step, and the data length of the eating motion pattern Y (n) is 80 for the time step. Therefore, the time series of the longitudinal acceleration GZ becomes the signal B, and the eating motion pattern Y (n) becomes the signal A. In the example shown in FIG. 22, the correlation coefficient is calculated while shifting the eating motion pattern Y (n) in increments of 1 hour with respect to the time series (signal B) of the longitudinal acceleration GZ (see FIG. 8). Therefore, in this case, as shown in FIG. 22C, 121 (= 200−80 + 1) correlation coefficients are calculated.

  In step 1822, the shape determination unit 158 determines whether or not the maximum value (shape similarity) of the correlation coefficient calculated in step 1818 is equal to or greater than the predetermined threshold Tz read in step 1806. If the maximum value of the correlation coefficient is equal to or greater than the predetermined threshold Tz, it is determined that the shape condition is satisfied (the shape is similar), and the process proceeds to Step 1900 of FIG. 19, and otherwise, the process proceeds to Step 1824.

  In step 1824, it is determined whether n is the last. If n is not the last, n = n + 1 and the process returns to step 1810. In this way, the shape determination process by the shape determination unit 158 is continued for all eating patterns (see FIG. 21B) until the maximum value of the correlation coefficient equal to or greater than the predetermined threshold Tz is obtained. The When n is the last (that is, when shape determination with all eating patterns is completed without obtaining the maximum value of the correlation coefficient equal to or greater than the predetermined threshold Tz), the process proceeds to step 1826.

  In step 1826, since the shape determination unit 158 does not satisfy the shape condition, the shape determination unit 158 determines that there is no eating operation in the window period related to the i-th window data. Thereby, window data related to other operations than the eating operation can be excluded.

  19, in step 1900, the X / Y-axis acceleration data acquisition unit 160 determines the vertical acceleration GX within the window period related to the i-th window data for which the maximum value of the correlation coefficient that is equal to or greater than the predetermined threshold Tz is obtained. Get the time series of. FIG. 23 is a diagram illustrating an example of a time-series acquisition range of the vertical acceleration GX. In FIG. 23, a window period related to the i-th window data in which the maximum value of the correlation coefficient that is equal to or greater than the predetermined threshold Tz among all the data in the observation acceleration data holding unit 180 is shown in a range 72. Yes. That is, the period from time step T (m) to time step T (m) +200 corresponds to the window period related to the i-th window data from which the maximum value of the correlation coefficient that is equal to or greater than the predetermined threshold Tz is obtained. In this case, the X / Y-axis acceleration data acquisition unit 160 acquires the time series 70 of the vertical acceleration GX within the period from the time step T (m) to the time step T (m) +200.

  In step 1902, the longitudinal operating condition determination unit 162 calculates a correlation coefficient between the time series of the longitudinal acceleration GZ input in step 1800 and the time series of the vertical acceleration GX acquired in step 1900. In this case, since the data length of the time series of the longitudinal acceleration GZ and the time series of the vertical acceleration GX are the same (constant window width W1), one correlation coefficient is calculated.

  In step 1904, the vertical direction operating condition determination unit 162 determines whether or not the correlation coefficient calculated in step 1902 is a positive value. If the correlation coefficient calculated in step 1902 is a positive value, the process proceeds to step 1908. Otherwise, the process proceeds to step 1906.

  In step 1906, the vertical operation condition determination unit 162 determines that there is no eating operation in the window period related to the i-th window data because the vertical operation condition is not satisfied. Thereby, it is possible to exclude window data related to operations other than the edible operation, which satisfy the amplitude condition and the shape condition in the Z-axis direction.

  In step 1908, the vertical direction operation condition determination unit 162 subtracts the minimum value of the vertical acceleration GX in the i-th window data from the maximum value of the vertical acceleration GX in the i-th window data, thereby obtaining the maximum and minimum difference in the vertical acceleration GX. Calculate amp.

  In step 1910, the vertical direction operation condition determination unit 162 determines whether the maximum / minimum difference amp of the vertical acceleration GX calculated in step 1908 is within a range of 500 mG to 4000 mG. The range of 500 mG to 4000 mG may correspond to a range that can be taken by the maximum / minimum difference amp of the time series of vertical acceleration actually obtained during the eating operation, and may be adapted by a test or the like. This numerical range is merely an example, and may be modified as appropriate. If the maximum / minimum difference amp of the vertical acceleration GX is in the range of 500 mG to 4000 mG, it is determined that the amplitude condition of the longitudinal operation condition is satisfied, and the process proceeds to step 1914. Otherwise, the process proceeds to step 1912.

  In step 1912, the vertical operation condition determination unit 162 determines that there is no edible movement in the window period related to the i-th window data because the vertical operation condition is not satisfied. Thereby, it is possible to exclude window data related to operations other than the edible operation, which satisfy the amplitude condition and the shape condition in the Z-axis direction.

  In step 1914, the lateral operation condition determination unit 164 reads the value of the predetermined threshold value Ty from the threshold data holding unit 184.

  In step 1916, the X / Y-axis acceleration data acquisition unit 160 obtains the time series of the lateral acceleration GY within the window period related to the i-th window data for which the maximum value of the correlation coefficient that is equal to or greater than the predetermined threshold Tz is obtained (see FIG. 23).

  In step 1918, the lateral operation condition determination unit 164 calculates the time-series variance value Vy of the left and right acceleration GY acquired in step 1916.

  In step 1920, the lateral operation condition determination unit 164 determines whether or not the variance value Vy calculated in step 1918 is smaller than the predetermined threshold value Ty read in step 1914. If the variance value Vy is smaller than the predetermined threshold value Ty, it is determined that the lateral operation condition is satisfied, and the process proceeds to step 1924. Otherwise, the process proceeds to step 1922.

  In step 1922, since the horizontal operation condition determination unit 164 does not satisfy the horizontal operation condition, the horizontal operation condition determination unit 164 determines that there is no eating operation in the window period related to the i-th window data. Thereby, it is possible to exclude the window data related to the operation other than the erosion operation and satisfy the amplitude condition and the shape condition in the Z-axis direction and the vertical operation condition in the X-axis direction.

  In step 1924, since the horizontal operation condition is satisfied, it is determined that there is an eating operation in the window period related to the i-th window data. As described above, according to the processing shown in FIGS. 18 and 19, only when the amplitude condition and shape condition in the Z-axis direction, the vertical operation condition in the X-axis direction, and the horizontal operation condition in the Y-axis direction are satisfied. , It is determined that there is an eating operation in the window period related to the i-th window data.

  Although each embodiment has been described in detail above, it is not limited to a specific embodiment, and various modifications and changes can be made within the scope described in the claims. It is also possible to combine all or a plurality of the components of the above-described embodiments.

  For example, in the above-described embodiment, the detection result of the eating motion is used for the detection of the meal period, but may be used for other purposes (for example, guidance for the time to take medicine after a meal). In this case, the eating motion frequency calculating unit 166, the meal period candidate detecting unit 168, the temporal validity determining unit 170, and the result holding unit 172 shown in FIG. 12 may be omitted, and correspondingly, the eating operation frequency calculating unit 166, the meal period candidate detecting unit 168, and the result holding unit 172 shown in FIG. In the processing, the processing after step 1310 may be omitted.

  In the embodiment described above, from the viewpoint of improving accuracy, in addition to the amplitude condition and the shape condition in the Z-axis direction, the vertical operation condition in the X-axis direction and the horizontal operation condition in the Y-axis direction are determined. Yes. However, determination of conditions other than the shape condition in the Z-axis direction may be omitted. In this case, it is not necessary to detect acceleration in the X-axis direction and the Y-axis direction. However, preferably, in addition to the shape condition in the Z-axis direction, the longitudinal operation condition in the X-axis direction (especially the condition regarding positive correlation) is determined. Other conditions may be added in any combination. Further, the determination order is arbitrary, and the determination of the shape condition in the Z-axis direction may be executed last.

  In the above-described embodiment, the acceleration data is divided into a plurality of windows using a window having a certain window width W1 in order to make it easy to handle acceleration data for a long time. However, such division may be omitted. In this case, the correlation coefficient may be calculated while shifting the eating motion pattern in predetermined time increments for all time series of acceleration data. In this case, since peak values of a plurality of correlation coefficients appear corresponding to a plurality of eating motions, the longitudinal operation condition in the X-axis direction is determined using acceleration data within a period in which the peak values of the correlation coefficients appear. Etc. may be determined.

  In the above-described embodiment, the eating action pattern held by the eating action pattern data holding unit 182 and various threshold values (predetermined threshold Tz etc.) held by the threshold value data holding unit 184 are externally set in advance (for example, the eating action pattern data holding unit 184). Generated by the designer of the motion detection device 1 and stored in the processing device 100. However, the eating action pattern and various threshold values may be automatically generated (learned) in the processing apparatus 100.

  In the above-described embodiment, the acceleration sensor 10 is mounted on the chest of the user. However, the acceleration sensor 10 may be mounted on another part (for example, the upper part of the abdomen) that moves substantially the same as the chest of the user. The upper part of the abdomen may be, for example, a part above the navel of the abdomen.

In addition, the following additional remarks are disclosed regarding the above Example.
(Appendix 1)
An acceleration sensor worn on the user's chest or on the top of the abdomen,
A eating motion detection device comprising: a processing device that determines whether or not there is a eating motion that is an operation in which the user carries food to a mouth based on an output signal of the acceleration sensor.
(Appendix 2)
The processing device determines that the user's eating action is present when a series of movements of the user such as “still, bent forward, still, back, rest” is detected based on an output signal of the acceleration sensor. The eating motion detection device according to appendix 1.
(Appendix 3)
The processing device determines the presence or absence of the user's eating operation based on a correlation coefficient between a time-series signal of a predetermined pattern corresponding to the series of operations and a time-series of output signals of the acceleration sensor. The eating motion detection device according to attachment 2.
(Appendix 4)
The time-series signal of the predetermined pattern is an output signal of the acceleration sensor generated during an operation other than the user's eating operation with respect to a time series of the output signal of the acceleration sensor generated during the user's eating operation. The eating motion detection device according to any one of appendices 1 to 3, which is a time-series signal that generates a correlation coefficient higher than that of the time series.
(Appendix 5)
The eating motion detection apparatus according to appendix 4, wherein a plurality of types of time-series signals having the predetermined pattern are prepared.
(Appendix 6)
The acceleration sensor detects longitudinal acceleration according to the longitudinal direction of the user,
The processing device determines that there is a eating action in the predetermined period when a correlation coefficient between the time series signal of the predetermined pattern and the time series of the longitudinal acceleration in a predetermined period is equal to or greater than a predetermined threshold. The eating motion detection device according to appendix 4 or 5.
(Appendix 7)
The processing apparatus further determines that there is a eating action in the predetermined period when the absolute value of the difference between the maximum value and the minimum value of the longitudinal acceleration in the predetermined period is within a predetermined range. The eating motion detection apparatus as described.
(Appendix 8)
The acceleration sensor detects longitudinal acceleration in the longitudinal direction of the user and vertical acceleration in the vertical direction of the user,
Of the appendices 1 to 7, the processing device determines the presence or absence of the user's eating action based on the relationship between the longitudinal acceleration according to the longitudinal direction of the user and the vertical acceleration according to the vertical direction of the user. The eating movement detection apparatus of any one of Claims.
(Appendix 9)
The acceleration sensor further detects vertical acceleration in the vertical direction of the user,
The processing device further includes a time series of the longitudinal acceleration in the predetermined period and the vertical acceleration in the predetermined period when the positive direction of the longitudinal acceleration is a backward direction and the positive direction of the vertical acceleration is an upward direction. The eating action detecting device according to appendix 6 or 7, wherein when there is a positive correlation with the time series, the eating action is determined to be present during the predetermined period.
(Appendix 10)
The processing device further determines that there is a eating operation in the predetermined period when the absolute value of the difference between the maximum value and the minimum value of the vertical acceleration in the predetermined period is within a predetermined range. The eating motion detection apparatus as described.
(Appendix 11)
The acceleration sensor further detects a lateral acceleration according to the lateral direction of the user,
11. The eating motion detection according to appendix 9 or 10, wherein the processing device further determines that there is a eating motion in the predetermined period when the degree of time-series variation of the lateral acceleration in the predetermined period is equal to or less than a predetermined value. apparatus.
(Appendix 12)
The processing device calculates a eating operation frequency per predetermined time based on the determination result of the presence or absence of the eating operation, and estimates a meal period based on the calculated eating operation frequency. The eating movement detection apparatus of any one of Claims.
(Appendix 13)
The eating operation detecting device according to appendix 12, wherein the processing device generates and outputs advice related to eating habits to a user based on the estimation result of the meal period.
(Appendix 14)
Obtain the output signal of the acceleration sensor worn on the chest of the user or on the upper part of the abdomen,
The eating motion detection method of determining the presence or absence of the eating motion which is an operation | movement which the said user carries food to a mouth based on the output signal of the said acceleration sensor.
(Appendix 15)
Obtain the output signal of the acceleration sensor worn on the chest of the user or on the upper part of the abdomen,
Based on the output signal of the acceleration sensor, the user determines the presence or absence of a eating action that is an action of bringing food to the mouth,
A program that causes a computer to execute processing.

DESCRIPTION OF SYMBOLS 1 Eating motion detection apparatus 2 Cloud 10 Acceleration sensor 100 Processing apparatus 152 Z-axis acceleration data acquisition part 154 Data division part 156 Amplitude condition determination part 158 Shape determination part 160 X / Y-axis acceleration data acquisition part 162 Longitudinal direction operation condition determination part 164 Lateral motion condition determining unit 166 Eating motion frequency calculating unit 168 Meal period candidate detecting unit 170 Temporal validity determining unit 172 Result holding unit 180 Observation acceleration data holding unit 182 Eating motion pattern data holding unit 184 Threshold data holding unit

Claims (7)

An acceleration sensor worn on the user's chest or on the top of the abdomen,
A processing device that determines the presence or absence of a eating action, which is an action of the user carrying food to the mouth, based on an output signal of the acceleration sensor ;
The processing device determines that the user's eating action is present when a series of movements of the user such as “still, bent forward, still, back, rest” is detected based on an output signal of the acceleration sensor. to, food operation detecting apparatus. The processing device determines the presence or absence of the user's eating operation based on a correlation coefficient between a time-series signal of a predetermined pattern corresponding to the series of operations and a time-series of output signals of the acceleration sensor. The eating movement detection device according to claim 1 . The acceleration sensor detects longitudinal acceleration according to the longitudinal direction of the user,
The processing device determines that there is a eating action in the predetermined period when a correlation coefficient between the time series signal of the predetermined pattern and the time series of the longitudinal acceleration in a predetermined period is equal to or greater than a predetermined threshold. The eating motion detection device according to claim 2 . The acceleration sensor further detects vertical acceleration in the vertical direction of the user,
The processing device further includes a time series of the longitudinal acceleration in the predetermined period and the vertical acceleration in the predetermined period when the positive direction of the longitudinal acceleration is a backward direction and the positive direction of the vertical acceleration is an upward direction. The eating motion detection device according to claim 3 , wherein when there is a positive correlation with the time series, the eating motion detection device determines that there is a eating motion in the predetermined period. The acceleration sensor further detects a lateral acceleration according to the lateral direction of the user,
5. The eating operation according to claim 3 , wherein the processing device further determines that there is an eating operation in the predetermined period when a time-series variation degree of the lateral acceleration in the predetermined period is equal to or less than a predetermined value. Detection device. Obtain the output signal of the acceleration sensor worn on the chest of the user or on the upper part of the abdomen,
Based on the output signal of the acceleration sensor, when the user detects a series of movements such as “still, bent forward, still, back, rest” , the eating action that the user carries food to the mouth A eating motion detection method executed by a computer, which is determined to be present . Obtain the output signal of the acceleration sensor worn on the chest of the user or on the upper part of the abdomen,
Based on the output signal of the acceleration sensor, when the user detects a series of movements such as “still, bent forward, still, back, rest” , the eating action that the user carries food to the mouth Judge that there is ,
A program that causes a computer to execute processing.

Images