Analysis device, method, and program

Abstract

To accurately analyze a cause of a load.SOLUTION: An analysis apparatus according to one embodiment includes a data acquisition unit, a state value calculation unit, a focused section setting unit, a clustering unit, and a stress information generation unit. The data acquisition unit acquires sensor data from a measurement target. The state value calculating unit calculates state values based on the sensor data. The focused section setting unit sets a plurality of focused sections in time series data based on the time series data of the state values and a predetermined criterion. The clustering unit performs clustering using the state values related to the plurality of focused sections and generates a clustering result. The stress information generating unit generates stress information including characteristic information of each of a plurality of clusters based on the clustering result.SELECTED DRAWING: Figure 2

Description

Technical Field

[0001] Embodiments of the present invention relate to an analytical apparatus, method, and program.

Background Art

[0002] With the spread of wearable devices equipped with sensors that measure acceleration, temperature and humidity, biological signals, etc., represented by activity meters and smartwatches, the development of technologies for recognizing a person's behavior and state using such wearable devices has been actively carried out. In recent years, there have also been active efforts to utilize them in industrial fields, such as classifying work content at work sites for analysis to improve productivity, or ensuring the safety of workers by performing fall detection and heat stroke risk estimation.

[0003] As one of such efforts, the development of technologies for measuring the load on workers by utilizing sensor data measured by wearable devices and images captured by cameras has also been actively carried out. Since work with excessive load causes work-related injuries such as worker injuries and productivity decline due to stress, the analysis of load in work has been pursued for a long time at work sites.

[0004] For example, the analysis of work load may sometimes be performed relying on on-site analysis by qualified experts. However, human analysis has many problems, such as not only taking a long time for analysis but also being difficult to analyze loads that are difficult to judge from appearance, such as mental stress. Therefore, automation of work load analysis using sensor data, images, etc. is expected to reduce the cost of analysis and be useful for analyzing loads that are difficult to judge from appearance.

[0005] In order to effectively reduce the load associated with work, it is essential to extract particularly high-load work from within a series of work sequences and also identify its causes. On the other hand, the load borne by a worker is influenced by various factors such as individual differences like differences in the build and physical strength of the worker, the shape of the work object, the working environment such as temperature and humidity, and the time of day of the work. Therefore, in order to accurately analyze the causes of the load, it is considered indispensable to analyze the differences in the way the load is imposed on the worker due to these factors.

[0006] As a method for evaluating the load associated with work, based on a model representing the human body, the pose is estimated from a video of the worker being filmed, the type of work is identified from the change pattern, and for each type of work, the pose bearing the load is determined and the load value applied to each part of the human body is calculated from its duration, thereby aggregating the load values for each type of work and each part. A technique is known.

[0007] However, although this technique can aggregate the loads for each type of work and each part of the human body, identify high-load work, and identify the parts bearing the load in that work, it cannot provide information for analyzing the differences in the way the load is generated due to factors such as the aforementioned individual differences and working environment. Therefore, there is a need for a technique that can accurately analyze the causes of the load from among many possible factors such as the individual differences of the worker and the working environment.

Prior Art Documents

Patent Documents

[0008]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0009] The problem to be solved by the present invention is to provide an analysis device, method, and program that can accurately analyze the causes of the load.

Means for Solving the Problem

[0010] An analyzer according to an embodiment includes a data acquisition unit, a state value calculation unit, a target interval setting unit, a clustering unit, and a stress information generation unit. The data acquisition unit acquires sensor data from a measurement target. The state value calculation unit calculates a state value based on the sensor data. The target interval setting unit sets a plurality of target intervals in the time series data based on the time series data of the state value and a predetermined reference. The clustering unit performs clustering using the state values related to the plurality of target intervals and generates a clustering result. The stress information generation unit generates stress information including feature information of each of the plurality of clusters based on the clustering result.

Brief Description of the Drawings

[0011]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Figure 11

Figure 12

Figure 13

Figure 14

Figure 15

Figure 16

Figure 17

Figure 18

Figure 19

Figure 20

Figure 21

Figure 22

Embodiments for Carrying Out the Invention

[0012] Hereinafter, embodiments of the analysis apparatus will be described in detail with reference to the drawings.

[0013] (First Embodiment) FIG. 1 is a block diagram illustrating the configuration of an analysis system including an analyzer according to a first embodiment. The analysis system 1 in FIG. 1 includes an analyzer 100, an output device 110, and one or more sensors. In FIG. 1, as the one or more sensors, a first sensor 121, a second sensor 122, and a third sensor 123 are exemplified. These three sensors each acquire sensor data for the same measurement target. The measurement target is, for example, a worker working at a work site such as a factory. The analyzer 100 analyzes information (stress information) representing the characteristics of stress related to the work based on the one or more sensor data. The output device 110 displays display data based on the stress information.

[0014] Note that the analysis system 1 may include sensors that acquire sensor data for other measurement targets. That is, the analyzer 100 may acquire one or more sensor data from a plurality of measurement targets.

[0015] Furthermore, the analysis system 1 may include a photographing device. The photographing device is, for example, a video camera (camera). The photographing device photographs, for example, a work place (for example, an assembly work place in a factory) where work is being performed by the measurement target, and acquires a still image or a moving image. In the present embodiment, the still image or the moving image acquired by the photographing device is referred to as a work video. Also, this work video may be used as sensor data.

[0016] Also, a plurality of photographing devices may be provided. The analysis system 1 may photograph the measurement target simultaneously with cameras installed at a plurality of positions to acquire a plurality of work videos in case the person to be measured is hidden by a work object or the like in a specific camera. Furthermore, the photographing device may be an infrared camera that can clearly photograph the silhouette of the measurement target even in work in a dark place.

[0017] The output device 110 is, for example, a monitor. The output device 110 receives display data from the analysis device 100. The output device 110 displays the display data. Note that the output device 110 is not limited to a monitor as long as it can display the display data. For example, the output device 110 may be a projector or a printer. Further, the output device 110 may include a speaker.

[0018] One or more sensors are incorporated, for example, in wearable devices worn on a plurality of body parts to be measured. Hereinafter, the wearable device and the sensor shall be used interchangeably. The plurality of body parts are, for example, the wrist, upper arm, ankle, thigh, waist, back, and head. Each of the one or more sensors acquires at least one measurement data such as acceleration data, angular velocity data, geomagnetic data, air pressure data, temperature and humidity data, myoelectric potential data, and pulse data as sensor data. The measurement data may include a plurality of channels. For example, when the measurement data is acceleration data, the sensor data includes data for three channels corresponding to the components in each direction of the acceleration (for example, the X-axis direction, the Y-axis direction, and the Z-axis direction). In the present embodiment, a case where the first sensor 121, the second sensor 122, and the third sensor 123 are used as the one or more sensors will be described.

[0019] The first sensor 121 is worn, for example, on the upper arm of the measurement target. The first sensor 121 measures the state of the upper arm of the measurement target as sensor data. The first sensor 121 outputs the measured sensor data to the analysis device 100.

[0020] The second sensor 122 is worn, for example, on the wrist of the measurement target. The second sensor 122 measures the state of the wrist of the measurement target as sensor data. The second sensor 122 outputs the measured sensor data to the analysis device 100.

[0021] The third sensor 123 is attached to, for example, the waist of the measurement target. The third sensor 123 measures the state of the waist of the measurement target as sensor data. The third sensor 123 outputs the measured sensor data to the analyzer 100.

[0022] In the following specific examples, it will be described assuming that the analyzer 100 uses the sensor data acquired from each of the first sensor 121, the second sensor 122, and the third sensor 123.

[0023] FIG. 2 is a block diagram illustrating the configuration of the analyzer according to the first embodiment. The analyzer 100 in FIG. 2 includes a data acquisition unit 210, a state value calculation unit 220, a target section setting unit 230, a clustering unit 240, a stress information generation unit 250, and a display control unit 260.

[0024] The data acquisition unit 210 acquires sensor data from each of the first sensor 121, the second sensor 122, and the third sensor 123. Hereinafter, when there is no need to distinguish these three pieces of sensor data, they are simply referred to as sensor data. The data acquisition unit 210 outputs the acquired sensor data to the state value calculation unit 220. The sensor data includes, for example, time and data in which the measurement data of each sensor is associated. Note that the data acquisition unit 210 may store the acquired sensor data in a storage unit (not shown in FIG. 2) provided in the analyzer 100, or may transmit it to an external storage device, a server, or the like.

[0025] The state value calculation unit 220 receives the sensor data from the data acquisition unit 210. The state value calculation unit 220 calculates a state value based on the sensor data. The state value calculation unit 220 outputs the time-series data of the calculated state value to the target section setting unit 230. Note that the state value calculation unit 220 may store the time-series data of the calculated state value in a storage unit provided in the analyzer 100, or may transmit it to an external storage device, a server, or the like.

[0026] The state value includes, for example, at least a physical stress value. The physical stress value includes, for example, a posture stress value that numerically represents the degree of load in a working posture that can cause stress (load) on a body part. The posture stress value is calculated, for example, based on sensor data, by estimating the angle and duration from a reference position at the site corresponding to the sensor attached to the measurement target, and based on the estimated angle and duration. The larger the value of the posture stress value, the greater the stress. Note that the posture stress value may also be referred to as a load value.

[0027] In addition, the state value may include a mental stress value. The mental stress value includes, for example, an LF / HF value that represents the frequency component of heart rate variability as a ratio. The LF / HF value is calculated by performing frequency analysis on the variation of the RR interval of the heart rate and taking the ratio of the power values of the low-frequency region (LF) and the high-frequency region (HF) of the spectrum. The smaller the LF / HF value, the more dominant the parasympathetic nerve, that is, it represents a relaxed state. The larger the LF / HF value, the more dominant the sympathetic nerve, that is, it represents a state of feeling stress.

[0028] Furthermore, the state value may include happiness. Happiness is estimated based on the amount of communication measured based on the voice data of the measurement target.

[0029] In the following specific examples, it is described that the state value is a posture stress value. Therefore, the time-series data of the state value is, for example, data in which time is associated with the posture stress value (load value) related to the site corresponding to each sensor.

[0030] The attention interval setting unit 230 receives the time-series data of the state value from the state value calculation unit 220. The attention interval setting unit 230 sets a plurality of attention intervals based on the time-series data of the state value and a predetermined reference. The attention interval setting unit 230 outputs information on the state value (attention interval information) regarding the set plurality of attention intervals to the clustering unit 240.

[0031] The predetermined criteria are, for example, a predetermined time length and a predetermined threshold value for the state value. Therefore, the target section setting unit 230 extracts, from the time series data of the state value, a section in which the state value exceeds the predetermined threshold value over a predetermined time length, and sets it as the target section.

[0032] Hereinafter, the time series data included in the target section will be referred to as a time series pattern. Further, when a plurality of time series patterns are included in the target section, these plurality of time series patterns will be referred to as a time series pattern set.

[0033] The clustering unit 240 receives the target section information from the target section setting unit 230. The clustering unit 240 performs clustering using the state values for the plurality of target sections included in the target section information, and generates a clustering result. The clustering unit 240 outputs the clustering result to the stress information generation unit 250.

[0034] Specifically, the clustering unit 240 generates a combined time series pattern by combining the time series patterns (time series pattern sets) for each of the plurality of elements included in the sensor data for each of the plurality of target sections. The plurality of elements indicate the parts corresponding to each sensor. For example, when sensors are attached to the upper arm, wrist, and waist of the measurement target, respectively, the clustering unit 240 generates a combined time series pattern by combining the time series pattern for the upper arm, the time series pattern for the wrist, and the time series pattern for the waist.

[0035] There are, for example, the following two methods for generating a composite time series pattern. In the first method, the clustering unit 240 generates a composite time series pattern by combining time series patterns (time series pattern sets) for each of a plurality of elements in the time direction. In the second method, the clustering unit 240 generates a composite time series pattern by superimposing time series patterns (time series pattern sets) for each of a plurality of elements. Note that the composite time series pattern generated by the first method may be called a combined time series pattern. Also, the order of combination may be arbitrary, but it is assumed to be unified among a plurality of time series pattern sets.

[0036] Next, the clustering unit 240 calculates the similarity between each of the generated plurality of composite time series patterns. The similarity is, for example, the distance between two time series patterns using the DTW (Dynamic Time Warping) method. The DTW method can evaluate the difference in the shape of time series patterns while considering stretching and shrinking in the time direction. Therefore, the clustering unit 240 can appropriately calculate the similarity even if the time lengths of the two time series patterns are different by using the DTW method.

[0037] Note that when calculating the similarity, the clustering unit 240 may align the time lengths of the two time series patterns by stretching and shrinking the time lengths of the time series patterns. When the time lengths of the two time series patterns are aligned, the clustering unit 240 may calculate not only the DTW method but also the Euclidean distance or the like as the similarity.

[0038] Furthermore, the clustering unit 240 performs clustering to group the target intervals with similar degrees of similarity as the same cluster based on the calculated multiple degrees of similarity, and generates a clustering result. As a clustering method, for example, the K-Means method (K-Means clustering) is used. Also, as a clustering method, any method such as hierarchical clustering, DBSCAN (Density-based spatial clustering of application with noise) may be used.

[0039] The clustering result includes, for example, data associating the information of the feature samples (e.g., sample ID) corresponding to the target intervals represented at positions on an arbitrary coordinate axis with the information for distinguishing each cluster (e.g., cluster ID).

[0040] Note that the clustering unit 240 does not necessarily have to generate a composite time series pattern. When not generating a composite time series pattern, the clustering unit 240 calculates the degree of similarity for each of the multiple time series patterns regarding the multiple elements in the target interval, and performs clustering based on the sum of the multiple degrees of similarity regarding the multiple elements. The calculation of the degree of similarity and the clustering may use the same method as the method described above.

[0041] The stress information generation unit 250 receives the clustering result from the clustering unit 240. The stress information generation unit 250 generates stress information including the feature information of each of the multiple clusters based on the clustering result. The stress information generation unit 250 outputs the generated stress information to the display control unit 260.

[0042] The characteristic information is, for example, a time-series pattern set (representative time-series pattern set) representing each cluster. The representative time-series pattern set may be an average time-series pattern set obtained by averaging a plurality of time-series pattern sets included in the cluster for each element, or may be a time-series pattern set of a target section where the similarity to the average time-series pattern set is maximized. Note that the average time-series pattern set may be calculated by at least either the clustering unit 240 or the stress information generation unit 250.

[0043] Furthermore, the stress information may include information related to the characteristic information (related information). The related information includes, for example, a work video corresponding to the time-series pattern set of the target section where the similarity to the average time-series pattern set is maximized and a graph representing the clustering result. Also, the related information may include a graph representing the time-series data of the state value.

[0044] The display control unit 260 receives the stress information from the stress information generation unit 250. The display control unit 260 generates display data based on the stress information and causes it to be displayed on the display which is the output device 110.

[0045] The configuration of the analysis system 1 and the analysis device 100 according to the first embodiment has been described above. Next, the operation of the analysis device 100 will be described with reference to the flowchart of FIG. 3.

[0046] FIG. 3 is a flowchart illustrating the operation of the analysis device according to the first embodiment. The processing of the flowchart in FIG. 3 starts when the analysis program is executed by the user.

[0047] (Step S310) When the analysis program is executed, the data acquisition unit 210 acquires sensor data from the measurement target.

[0048] (Step S320) After the sensor data is acquired, the state value calculation unit 220 calculates a state value based on the sensor data. Hereinafter, FIG. 4 will be used to explain an example of the time series data of the calculated state value.

[0049] FIG. 4 is a graph showing the load values of the upper arm, wrist, and waist of an operator over time in the first embodiment. In FIG. 4, a graph 410 regarding the load value of the upper arm, a graph 420 regarding the load value of the wrist, and a graph 430 regarding the load value of the waist are shown on a common time axis. These graphs 410, 420, and 430 respectively correspond to the time series data of the state value.

[0050] (Step S330) After the state value is calculated, the target section setting unit 230 sets a plurality of target sections based on the time series data of the state value and a predetermined reference. Hereinafter, FIG. 4 will be used again to explain a specific example of setting a plurality of target sections.

[0051] First, the graph 410 will be described. The target section setting unit 230 sets a threshold value th1 regarding the load value of the upper arm, and extracts, as a target section, a section in the graph 410 that exceeds the threshold value th1 for a predetermined time length or more. In FIG. 4, as the extracted target sections, three target sections, i.e., a target section s1 from time t1 to time t2, a target section s3 from time t5 to time t6, and a target section s4 from time t7 to time t8, are shown.

[0052] Next, the graph 420 will be described. The target section setting unit 230 sets a threshold value th2 regarding the load value of the wrist, and extracts, as a target section, a section in the graph 420 that exceeds the threshold value th2 for a predetermined time length or more. In FIG. 4, as the extracted target sections, three target sections, i.e., a target section s2 from time t3 to time t4, a target section s5 from time t9 to time t10, and a target section s6 from time t11 to time t12, are shown.

[0053] Finally, the graph 430 will be described. The target section setting unit 230 sets a threshold value th3 for the lumbar load value, and extracts, as a target section, a section in the graph 430 that exceeds the threshold value th3 for a predetermined time period or longer. Note that in FIG. 4, no target section has been extracted for the graph 430.

[0054] Here, the above-described terms will be reconfirmed with reference to FIG. 4. For example, the time-series data of the graph 410 included in the target section s1 is called a time-series pattern. Similarly, the time-series data of the graph 420 and the time-series data of the graph 430 included in the target section s1 are also each called a time-series pattern. Then, a set of these multiple time-series patterns is called a time-series pattern set. Therefore, the time-series pattern set of the target section s1 includes a time-series pattern related to the upper arm, a time-series pattern related to the wrist, and a time-series pattern related to the waist included in the target section s1 from time t1 to time t2. The same applies to other target sections.

[0055] Note that the target section setting unit 230 may relax the conditions (setting criteria) for setting the target section. For example, when the load value in the time-series data is below the threshold value, if the time length of the section where it is below is smaller than a predetermined value, the target sections before and after that section may be combined and regarded as a target section when the combined time length exceeds a predetermined time length. Hereinafter, other examples of the setting criteria for multiple target sections will be described with reference to FIGS. 5 and 6.

[0056] FIG. 5 is a diagram for explaining another example of the setting criteria for the target section in the first embodiment. In FIG. 5, a graph 510 related to the load value of the upper arm is shown. The target section setting unit 230 sets a threshold value th1 for the load value of the upper arm, and extracts, as a target section, a section in the graph 510 that exceeds the threshold value th1 for a predetermined time period or longer.

[0057] For example, in FIG. 5, for graph 510, in the small interval ss21 from time t21 to time t22, it exceeds the threshold th1, and in the gap interval gs from time t22 to time t23, it temporarily falls below the threshold th1, and in the small interval ss22 from time t23 to time t24, it exceeds the threshold th1. Here, since both the small interval ss21 and the small interval ss22 are shorter than a predetermined time length based on a predetermined criterion, it is assumed that they do not meet the criterion as the target interval. Also, the gap interval gs is assumed to be smaller than a predetermined value. At this time, the target interval setting unit 230 ignores the gap interval gs and compares the interval from time t21 to time t24 with a predetermined time length. As a result, the target interval shown in FIG. 6 below is set.

[0058] FIG. 6 is a diagram for explaining the target interval set in another example of FIG. 5. In FIG. 6, the graph 510 is shown in the same manner as in FIG. 5. In FIG. 6, it is assumed that the interval from time t21 to time t24, that is, the interval obtained by adding the small interval ss21, the gap interval gs, and the small interval ss22 together, exceeds a predetermined time length based on a predetermined criterion. Therefore, the target interval setting unit 230 sets the interval from time t21 to time t24 as the target interval s21.

[0059] (Step S340) After a plurality of target intervals are set, the clustering unit 240 performs clustering using the state values related to the plurality of target intervals and generates a clustering result. Hereinafter, the process of step S340 will be referred to as "clustering process". A specific example of the clustering process will be described using the flowchart of FIG. 7.

[0060] FIG. 7 is a flowchart illustrating the clustering process of the flowchart of 3. The flowchart of FIG. 7 transitions from step S330 of FIG. 3. Note that in the flowchart of FIG. 7, an example will be described in which the clustering unit 240 generates a synthesized time series pattern (combined time series pattern) by combining time series pattern sets in the time direction.

[0061] (Step S341) After a plurality of target intervals are set, the clustering unit 240 combines the time-series patterns of each part for each of the plurality of target intervals. Hereinafter, the process of combining time-series patterns (combination process) will be described with reference to FIG. 8.

[0062] FIG. 8 is a diagram for explaining the combination process in the first embodiment. In FIG. 8, a time-series pattern set of the target interval s1 is shown. The time-series pattern set includes the time-series data 810 of the upper arm, the time-series data 820 of the wrist, and the time-series data 830 of the waist. The clustering unit 240 generates a combined time-series pattern cp1 by performing a combination process 800 on the time-series pattern set of the target interval s1. The combined time-series pattern cp1 corresponds to a graph 840 obtained by combining the time-series data 810, the time-series data 820, and the time-series data 830 in the time direction. Note that the combination process 800 is similarly performed for the target intervals s2 to s6 shown in FIG. 4.

[0063] (Step S342) After combining the time-series patterns, the clustering unit 240 calculates the similarity of each of the plurality of combined time-series patterns. Hereinafter, the process of calculating the similarity of each of the plurality of time-series patterns (similarity calculation process) will be described with reference to FIG. 9.

[0064] FIG. 9 is a diagram for explaining the similarity calculation process in the first embodiment. In FIG. 9, a combined time-series pattern cp1 corresponding to the target interval s1 and a combined time-series pattern cp3 corresponding to the target interval s3 are shown. The clustering unit 240 calculates the similarity by performing a similarity calculation process 900 based on the combined time-series pattern cp1 and the combined time-series pattern cp3. Note that the similarity calculation process 900 is performed for all different two combinations of the plurality of combined time-series patterns corresponding to the plurality of target intervals shown in FIG. 4.

[0065] (Step S343) After calculating the similarity, the clustering unit 240 performs clustering based on the calculated multiple similarities and generates a clustering result. After step S343, the process proceeds to step S350 in FIG. 3. Hereinafter, the clustering result will be described with reference to FIG. 10.

[0066] FIG. 10 is a diagram for explaining the clustering result of a plurality of feature samples corresponding to a plurality of target sections plotted on two-dimensional coordinates in the first embodiment. In FIG. 10, the horizontal axis in the two-dimensional coordinates corresponds to the wrist load value, and the vertical axis corresponds to the upper arm load value. In the clustering result 1000 of FIG. 10, as two clusters, a first cluster cl1 and a second cluster cl2 are shown. The first cluster cl1 includes a feature sample c1 corresponding to the target section s1, a feature sample c3 corresponding to the target section s3, and a feature sample c4 corresponding to the target section s4. The second cluster cl2 includes a feature sample c2 corresponding to the target section s2, a feature sample c5 corresponding to the target section s5, and a feature sample c6 corresponding to the target section s6.

[0067] (Step S350) After the clustering result is generated, the stress information generation unit 250 generates stress information including the feature information of each of the plurality of clusters based on the clustering result. Hereinafter, an example will be described in which the feature information is a time series pattern set of the target section with the maximum similarity to the average time series pattern set. The average time series pattern set corresponds to an average feature sample that is the average of the feature samples in each cluster. The average feature sample will be described with reference to FIG. 11.

[0068] FIG. 11 is a diagram for explaining the average feature samples in each cluster of FIG. 10. FIG. 11 shows a first average feature sample ave1 which is the average of the feature samples included in the first cluster cl1, and a second average feature sample ave2 which is the average of the feature samples included in the second cluster cl2. The stress information generation unit 250 determines the above two average feature samples by aggregating and averaging the feature samples included in each cluster.

[0069] Furthermore, the stress information generation unit 250 determines, for each cluster, a feature sample close to the average feature sample. Hereinafter, it is assumed that the stress information generation unit 250 determines a feature sample c1 close to the first average feature sample ave1 for the first cluster cl1, and determines a feature sample c6 close to the second average feature sample ave2 for the second cluster cl2. Therefore, the stress information generation unit 250 generates, as feature information, a time series pattern set of the target section s1 corresponding to the feature sample c1 and a time series pattern set of the target section s6 corresponding to the feature sample c6. At this time, the feature information (time series pattern set) is generated, for example, by extracting from the time series data of the state values shown in FIG. 4.

[0070] Then, the stress information generation unit 250 generates stress information including the generated feature information and related information regarding the feature information.

[0071] (Step S360) After the stress information is generated, the display control unit 260 causes the stress information to be displayed. Specifically, the display control unit 260 causes the display data based on the stress information to be displayed on the display which is the output device 110. After step S360, the analysis program ends. Hereinafter, an example of the display data in the case where the work video is included as the related information will be described with reference to FIGS. 12 and 13.

[0072] FIG. 12 is a diagram illustrating display data based on stress information in the first embodiment. The display data 1200 in FIG. 12 includes a display area 1210 for displaying information related to the first cluster and a display area 1220 for displaying information related to the second cluster.

[0073] In the display area 1210, a work video 1211 representing the first cluster and a time-series pattern set 1212 are displayed. The work video 1211 is a video corresponding to the time-series pattern set 1212. The time-series pattern set 1212 is a time-series pattern set of the target section s1 corresponding to the feature sample c1 in FIGS. 10 and 11.

[0074] In the display area 1220, a work video 1221 representing the second cluster and a time-series pattern set 1222 are displayed. The work video 1221 is a video corresponding to the time-series pattern set 1222. The time-series pattern set 1222 is a time-series pattern set of the target section s6 corresponding to the feature sample c6 in FIGS. 10 and 11.

[0075] Therefore, according to the display data 1200, since it is grouped according to the difference in load in the same work, the user can confirm the difference in load in the same work by viewing the display data 1200. For example, in the time-series pattern set 1212, the time-series pattern of the upper arm exceeds the threshold value, and the work video 1211 shows that a load is applied to the upper arm. On the other hand, in the time-series pattern set 1222, the time-series pattern of the wrist exceeds the threshold value, and the work video 1221 shows that a load is applied to the wrist.

[0076] FIG. 13 is a diagram illustrating other display data based on stress information in the first embodiment. The display data 1300 in FIG. 13 has a display area 1310 for displaying information about the first cluster and a display area 1320 for displaying information about the second cluster. The display data 1300 is different from the display data 1200 in FIG. 12 in that it displays all the work videos and time series pattern sets of each cluster.

[0077] In the display area 1310, work videos 1311, 1313, 1315 in the first cluster and time series pattern sets 1312, 1314, 1316 are displayed. The work videos 1311, 1313, 1315 are videos corresponding to the time series pattern sets 1312, 1314, 1316, respectively.

[0078] The time series pattern set 1312 is a time series pattern set of the target section s1 corresponding to the feature sample c1 in FIGS. 10 and 11. Also, the time series pattern set 1314 is a time series pattern set of the target section s3 corresponding to the feature sample c3 in FIGS. 10 and 11. Also, the time series pattern set 1316 is a time series pattern set of the target section s4 corresponding to the feature sample c4 in FIGS. 10 and 11.

[0079] In the display area 1320, work videos 1321, 1323, 1325 in the second cluster and time series pattern sets 1322, 1324, 1326 are displayed. The work videos 1321, 1323, 1325 are videos corresponding to the time series pattern sets 1322, 1324, 1326, respectively.

[0080] The time series pattern set 1322 is a time series pattern set of the target section s2 corresponding to the feature sample c2 in FIGS. 10 and 11. Also, the time series pattern set 1324 is a time series pattern set of the target section s5 corresponding to the feature sample c5 in FIGS. 10 and 11. Also, the time series pattern set 1326 is a time series pattern set of the target section s6 corresponding to the feature sample c6 in FIGS. 10 and 11.

[0081] Therefore, according to the display data 1300, all the data belonging to each cluster are displayed in a list. Thus, by visually recognizing the display data 1300, the user can confirm the difference in load in the same operation from an overview perspective. Also, when focusing on each cluster, the user can check the operations classified as the same type of load in a list. Thereby, for example, even when there is a discrepancy between the apparent operation and the actual load (for example, when the upper arm seems to be loaded in the work video but actually the wrist is loaded), the user can correctly analyze the cause of the load.

[0082] Note that the display data 1200 and the display data 1300 display information about two clusters, but it is not limited to this. For example, when the clustering result includes three or more clusters, the display data may display information about three or more clusters. Also, for example, the display data may select and display information about a specific cluster among a plurality of clusters. Hereinafter, with respect to the clustering result in FIG. 10, FIGS. 14 and 15 will be used to explain an example of displaying information about the first cluster, and FIGS. 16 and 17 will be used to explain an example of displaying information about the second cluster.

[0083] FIG. 14 is a diagram illustrating display data regarding the first cluster in the first embodiment. The display data 1400 in FIG. 14 includes a graph 1410 of time-series data of state values, a clustering result 1420, and a work video 1430.

[0084] Graph 1410 includes graph 410, graph 420, and graph 430 in FIG. 4. Also, for six regions of interest in graph 1410, the first cluster and the second cluster are color-coded in a distinguishable manner. This color-coding corresponds to the color-coding of the clustering in clustering result 1420 described later. Also, in graph 1410, time series pattern set 1411 corresponding to region of interest s1 representing the first cluster is emphasized and displayed.

[0085] Clustering result 1420 is substantially the same as clustering result 1000 in FIG. 10. In clustering result 1420, different color-codings are used for the respective ranges of the first cluster cl1 and the second cluster cl2. Also, in clustering result 1420, feature sample 1421 corresponding to region of interest s1 representing the first cluster is emphasized and displayed.

[0086] Work video 1430 is the same as work video 1211 in FIG. 12. Work video 1430 corresponds to time series pattern set 1411 in graph 1410 and feature sample 1421 in clustering result 1420, respectively.

[0087] Therefore, according to display data 1400, since time series pattern set 1411 representing the first cluster can be displayed in association with clustering result 1420 and work video 1430, the user can immediately check the outline of the first cluster by visually recognizing display data 1400. Note that the same display can be performed for other time series pattern sets included in the first cluster.

[0088] FIG. 15 is a diagram illustrating other display data related to the first cluster in the first embodiment. Display data 1500 in FIG. 15 includes graph 1510 for time series data of state values and a plurality of work videos 1521, 1522, 1523.

[0089] Graph 1510 is substantially the same as graph 1410. The difference from graph 1410 is that in graph 1510, a plurality of time series pattern sets 1511, 1512, 1513 corresponding to a plurality of target intervals s1, s3, s4 belonging to the first cluster are emphasized and displayed.

[0090] The work video 1521 is the same as the work video 1311 in FIG. 13. The work video 1521 corresponds to the time series pattern set 1511 in graph 1510. The combination of the work video 1521 and the time series pattern set 1511 is the same as the combination of the work video 1311 and the time series pattern set 1312 in FIG. 13.

[0091] The work video 1522 is the same as the work video 1313 in FIG. 13. The work video 1522 corresponds to the time series pattern set 1512 in graph 1510. The combination of the work video 1522 and the time series pattern set 1512 is the same as the combination of the work video 1313 and the time series pattern set 1314 in FIG. 13.

[0092] The work video 1523 is the same as the work video 1315 in FIG. 13. The work video 1523 corresponds to the time series pattern set 1513 in graph 1510. The combination of the work video 1523 and the time series pattern set 1513 is the same as the combination of the work video 1315 and the time series pattern set 1316 in FIG. 13.

[0093] Therefore, according to the display data 1500, since a plurality of time series pattern sets 1511, 1512, 1513 related to the first cluster and a plurality of work videos 1521, 1522, 1523 can be associated and displayed, the user can check the information included in the first cluster at a glance by visually recognizing the display data 1500.

[0094] FIG. 16 is a diagram illustrating display data related to the second cluster in the first embodiment. The display data 1600 in FIG. 16 includes a graph 1610 for time-series data of state values, a clustering result 1620, and a work video 1430.

[0095] The graph 1610 is substantially the same as the graph 1410. Different from the graph 1410, in the graph 1610, a time-series pattern set 1611 corresponding to the attention section s6 representing the second cluster is emphasized and displayed.

[0096] The clustering result 1620 is substantially the same as the clustering result 1420. Different from the clustering result 1420, in the clustering result 1620, a feature sample 1621 corresponding to the attention section s6 representing the second cluster is emphasized and displayed.

[0097] The work video 1630 is the same as the work video 1221 in FIG. 12. The work video 1221 corresponds to the time-series pattern set 1611 in the graph 1610 and the feature sample 1621 in the clustering result 1620, respectively.

[0098] Therefore, according to the display data 1600, since the time-series pattern set 1611 representing the second cluster can be displayed in association with the clustering result 1620 and the work video 1630, the user can immediately confirm the outline of the second cluster by visually recognizing the display data 1600. Note that the same display can be performed for other time-series pattern sets included in the second cluster.

[0099] FIG. 17 is a diagram illustrating other display data related to the second cluster in the first embodiment. The display data 1700 in FIG. 17 includes a graph 1710 for time-series data of state values and a plurality of work videos 1721, 1722, 1723.

[0100] Graph 1710 is substantially the same as Graph 1410. The difference from Graph 1410 is that in Graph 1710, a plurality of time series pattern sets 1711, 1712, 1713 corresponding to a plurality of target intervals s2, s5, s6 belonging to the second cluster are emphasized and displayed.

[0101] The work video 1721 is the same as the work video 1321 in FIG. 13. The work video 1721 corresponds to the time series pattern set 1711 in Graph 1710. The combination of the work video 1721 and the time series pattern set 1711 is the same as the combination of the work video 1321 and the time series pattern set 1322 in FIG. 13.

[0102] The work video 1722 is the same as the work video 1323 in FIG. 13. The work video 1722 corresponds to the time series pattern set 1712 in Graph 1710. The combination of the work video 1722 and the time series pattern set 1712 is the same as the combination of the work video 1323 and the time series pattern set 1324 in FIG. 13.

[0103] The work video 1723 is the same as the work video 1325 in FIG. 13. The work video 1723 corresponds to the time series pattern set 1713 in Graph 1710. The combination of the work video 1723 and the time series pattern set 1713 is the same as the combination of the work video 1325 and the time series pattern set 1326 in FIG. 13.

[0104] Therefore, according to the display data 1700, since a plurality of time series pattern sets 1711, 1712, 1713 related to the second cluster and a plurality of work videos 1721, 1722, 1723 can be associated and displayed, the user can view the display data 1700 to check the information included in the second cluster at a glance.

[0105] As described above, according to the display data 1400, display data 1500, display data 1600, and display data 1700, by visually recognizing these display data, the user can visually confirm the content of the work for each cluster, and can easily perform a load factor analysis.

[0106] As described above, the analysis device according to the first embodiment acquires sensor data from a measurement target, calculates a state value based on the sensor data, sets a plurality of target intervals in the time-series data based on the time-series data of the state value and a predetermined reference, performs clustering using the state values for the plurality of target intervals, generates a clustering result, and generates stress information including feature information for each of the plurality of clusters based on the clustering result.

[0107] Therefore, the analysis device according to the first embodiment can classify the state of the measurement target by performing clustering using a plurality of target intervals, and thus can accurately analyze the cause of the load.

[0108] (Application Example of the First Embodiment) In the first embodiment, the comparison of one threshold value and the load value when setting the target interval has been described. On the other hand, in the application example of the first embodiment, the comparison of a plurality of threshold values and the load value when setting the target interval will be described.

[0109] FIG. 18 is a diagram for explaining the setting criteria of candidate target intervals in the application example of the first embodiment. In FIG. 18, a graph 1810 regarding the load value of the upper arm is shown. The target interval setting unit 230 sets a first threshold value th11 regarding the load value of the upper arm and a second threshold value th12 that is equal to or less than the first threshold value th11. After setting the two threshold values, the target interval setting unit 230 extracts, as the target interval, an interval that exceeds the first threshold value th11 over a predetermined time period in the graph 1810, and extracts, as the candidate target interval, an interval that exceeds the second threshold value th12.

[0110] For example, in FIG. 18, for graph 1810, it exceeds the first threshold th11 in the interval from time t32 to time t33, and exceeds the second threshold th12 in the interval from time t31 to time t34. Therefore, the target interval setting unit 230 sets the interval from time t32 to time t33 as the target interval s31, and sets the interval from time t31 to time t34 as the candidate target interval cs31.

[0111] In FIG. 18, the target interval s31 is included in the candidate target interval cs31. However, when the load value changes between the first threshold th11 and the second threshold th12 or more, it is conceivable that only the candidate target interval is set. Therefore, the subsequent clustering process may use the candidate target interval instead of the target interval.

[0112] (Second Embodiment) In the first embodiment and the application example of the first embodiment, the case where the type of the operator's operation (operation type) is not considered or the operation type is known has been described. On the other hand, in the second embodiment, the case of determining the operation type of the operator from the sensor data will be described.

[0113] FIG. 19 is a block diagram illustrating the configuration of the analyzer according to the second embodiment. The analyzer 1900 in FIG. 19 includes a data acquisition unit 1910, a state value calculation unit 1920, a target interval setting unit 1930, a clustering unit 1940, a stress information generation unit 1950, a display control unit 1960, and an operation type determination unit 1970.

[0114] Note that the state value calculation unit 1920, the target interval setting unit 1930, the stress information generation unit 1950, and the display control unit 1960 have the same operations as the state value calculation unit 220, the target interval setting unit 230, the stress information generation unit 250, and the display control unit 260 in FIG. 2, and thus the description thereof is omitted.

[0115] The data acquisition unit 1910 acquires sensor data from each of the first sensor 121, the second sensor 122, and the third sensor 123. The data acquisition unit 1910 outputs the acquired sensor data to the state value calculation unit 1920 and the operation type determination unit 1970.

[0116] The operation type determination unit 1970 receives sensor data from the data acquisition unit 1910. The operation type determination unit 1970 determines the operation type based on the sensor data. The operation type determination unit 1970 outputs information on the determined operation type (operation type information) to the clustering unit 1940.

[0117] The clustering unit 1940 receives the attention interval information from the attention interval setting unit 1930 and receives the operation type information from the operation type determination unit 1970. The clustering unit 1940 performs clustering using the state values and operation types related to a plurality of attention intervals, and generates a clustering result. The clustering unit 1940 outputs the clustering result to the stress information generation unit 1950.

[0118] Specifically, the clustering unit 1940 classifies a plurality of attention intervals for each operation type. Next, the clustering unit 1940 generates a synthesized time series pattern for each of the plurality of attention intervals for each operation type by synthesizing time series pattern sets. Next, the clustering unit 1940 calculates the similarity between the generated plurality of synthesized time series patterns for each operation type. Finally, the clustering unit 1940 performs clustering for each operation type based on the calculated plurality of similarities, and generates a clustering result.

[0119] The configuration of the analysis device 1900 according to the second embodiment has been described above. Next, the operation of the analysis device 1900 will be described with reference to the flowchart of FIG. 20.

[0120] FIG. 20 is a flowchart illustrating the operation of the analysis device according to the second embodiment. The processing of the flowchart in FIG. 20 starts when the analysis program is executed by the user.

[0121] (Step S2010) When the analysis program is executed, the data acquisition unit 1910 acquires sensor data from the measurement target.

[0122] (Step S2020) After the sensor data is acquired, the state value calculation unit 1920 calculates a state value based on the sensor data.

[0123] (Step S2030) After the state value is calculated, the target section setting unit 1930 sets a plurality of target sections based on the time series data of the state value and a predetermined reference.

[0124] (Step S2040) After a plurality of target sections are set, the operation type determination unit 1970 determines the operation type based on the sensor data.

[0125] (Step S2050) After the operation type is determined, the clustering unit 1940 performs clustering using the state values and the operation type related to the plurality of target sections, and generates a clustering result. Hereinafter, the process of step S2050 is referred to as "clustering process". A specific example of the clustering process will be described with reference to the flowchart of FIG. 21.

[0126] FIG. 21 is a flowchart illustrating the clustering process of the flowchart of FIG. 20. The flowchart of FIG. 21 transitions from step S2040 of FIG. 20.

[0127] (Step S2051) After the operation type is determined, the clustering unit 1940 classifies the plurality of target sections for each operation type.

[0128] (Step S2052) After classifying the plurality of target sections into a plurality of operation types, the clustering unit 1940 combines the time series patterns of each part for each of the plurality of target sections for each operation type.

[0129] (Step S2053) After combining the time series patterns, the clustering unit 1940 calculates the similarity of each of the plurality of combined time series patterns for each operation type.

[0130] (Step S2054) After calculating the similarity, the clustering unit 1940 performs clustering for each operation type based on the calculated plurality of similarities, and generates a clustering result. After step S2054, the process proceeds to step S2060.

[0131] (Step S2060) After the clustering result is generated, the stress information generation unit 1950 generates stress information including the feature information of each of the plurality of clusters based on the clustering result.

[0132] (Step S2070) After the stress information is generated, the display control unit 1960 causes the stress information to be displayed. Specifically, the display control unit 1960 causes the display data based on the stress information to be displayed on the display which is the output device 110. After step S2070, the analysis program ends.

[0133] In the second embodiment, since clustering processing considering the operation type is performed, the display data may display various information for each of the plurality of operation types. Further, the display data may display various information so that the plurality of operation types can be compared.

[0134] As described above, the analyzer according to the second embodiment acquires sensor data from a measurement target, calculates a state value based on the sensor data, determines the operation type of the measurement target based on the sensor data, sets a plurality of target intervals in the time-series data based on the time-series data of the state value and a predetermined reference, performs clustering using the state values for the plurality of target intervals for each operation type, generates a clustering result, and generates stress information including feature information for each of the plurality of clusters based on the clustering result.

[0135] Therefore, the analyzer according to the second embodiment can classify the state of the measurement target for each operation type by determining the operation type from the work of the measurement target, and thus can more accurately analyze the cause of the load.

[0136] Therefore, the analyzer according to the first embodiment can classify the state of the measurement target in detail by performing clustering using a plurality of target intervals, and thus can accurately analyze the cause of the load.

[0137] (Hardware Configuration) FIG. 22 is a block diagram illustrating the hardware configuration of a computer according to an embodiment. The computer 2200 includes, as hardware, a CPU (Central Processing Unit) 2210, a RAM (Random Access Memory) 2220, a program memory 2230, an auxiliary storage device 2240, and an input / output interface 2250. The CPU 2210 communicates with the RAM 2220, the program memory 2230, the auxiliary storage device 2240, and the input / output interface 2250 via a bus 2260.

[0138] The CPU 2210 is an example of a general-purpose processor. The RAM 2220 is used by the CPU 2210 as a working memory. The RAM 2220 includes a volatile memory such as SDRAM (Synchronous Dynamic Random Access Memory). The program memory 2230 stores various programs including the analysis program. As the program memory 2230, for example, a ROM (Read-Only Memory), a part of the auxiliary storage device 2240, or a combination thereof is used. The auxiliary storage device 2240 stores data non-temporarily. The auxiliary storage device 2240 includes a non-volatile memory such as an HDD or an SSD.

[0139] The input / output interface 2250 is an interface for connecting to or communicating with other devices. The input / output interface 2250 is used, for example, for connection to or communication with the output device 110, the first sensor 121, the second sensor 122, and the third sensor 123 shown in FIG. 1.

[0140] Each program stored in the program memory 2230 includes computer-executable instructions. When the program (computer-executable instructions) is executed by the CPU 2210, it causes the CPU 2210 to execute a predetermined process. For example, when the load estimation program is executed by the CPU 2210, it causes the CPU 2210 to execute a series of processes described with respect to each part of FIGS. 3, 7, 20, and 21.

[0141] The program may be provided to the computer 2200 in a state stored in a computer-readable storage medium. In this case, for example, the computer 2200 further includes a drive (not shown) for reading data from the storage medium, and acquires the program from the storage medium. Examples of the storage medium include magnetic disks, optical disks (such as CD-ROM, CD-R, DVD-ROM, DVD-R), magneto-optical disks (such as MO), and semiconductor memories. Also, the program may be stored in a server on a communication network, and the computer 2200 may download the program from the server using the input / output interface 2250.

[0142] The processing described in the embodiments is not limited to being performed by a general-purpose hardware processor such as the CPU 2210 executing a program, and may be performed by a dedicated hardware processor such as an ASIC (Application Specific Integrated Circuit). The term processing circuit (processing unit) includes at least one general-purpose hardware processor, at least one dedicated hardware processor, or a combination of at least one general-purpose hardware processor and at least one dedicated hardware processor. In the example shown in FIG. 22, the CPU 2210, the RAM 2220, and the program memory 2230 correspond to the processing circuit.

[0143] Therefore, according to each of the above embodiments, the cause of the load can be accurately analyzed.

[0144] Although some embodiments of the present invention have been described, these embodiments are presented as examples and are not intended to limit the scope of the invention. These novel embodiments can be implemented in various other forms, and various omissions, replacements, and changes can be made without departing from the gist of the invention. These embodiments and their modifications are included in the scope and gist of the invention, and are included in the invention described in the claims and its equivalent scope.

Description of Reference Numerals

[0145] 1… analysis system, 100, 1900… analysis device, 110… output device, 121… first sensor, 122… second sensor, 123… third sensor, 210, 1910… data acquisition unit, 220, 1920… state value calculation unit, 230, 1930… target interval setting unit, 240, 1940… clustering unit, 250, 1950… stress information generation unit, 260, 1960… display control unit, 410, 420, 430, 510, 840, 1410, 1510, 1610, 1710, 1810… graph, 800… combination process, 810, 820, 830… time series data, 900… similarity calculation process, 1000, 1420, 1620… clustering result, 1200, 1300, 1400, 1500, 1600, 1700… display data, 1210, 1220, 1310, 1320… display area, 1211, 1221, 1311, 1313, 1315, 1321, 1323, 1325, 1430, 1521, 1522, 1523, 1630, 1721, 1722, 1723… work video, 1212, 1222, 1312, 1314, 1316, 1322, 1324, 1326, 1411, 1511, 1512, 1513, 1611, 1711, 1712, 1713… time series pattern set, 1421, 1621… feature sample, 1970… operation type determination unit, 2200… computer, 2210… CPU, 2220… RAM, 2230… program memory, 2240… auxiliary storage device, 2250… input / output interface, 2260… bus, ave1… first average feature sample, ave2… second average feature sample, c1, c2, c3, c4, c5, c6… feature sample, cl1… first cluster, cl2… second cluster, cp1, cp3… combined time series pattern, cs31… candidate target interval, gs… gap interval, s1, s2, s3, s4, s5, s6, s21, s31… target interval, ss21, ss22… small interval, th1, th2, th3… threshold, th11… first threshold, th12… second threshold.

Claims

1. A data acquisition unit that acquires sensor data from a measurement target, A state value calculation unit that calculates a state value based on the sensor data, An attention interval setting unit that sets a plurality of attention intervals in the time series data based on the time series data of the state value and a predetermined reference, A clustering unit that performs clustering using the time series data of the state value for the plurality of attention intervals and generates a clustering result, A generation unit that generates display information including feature information for each of the plurality of clusters based on the clustering result, Comprising, The state value includes a load value that numerically represents the degree of load in a working posture that can cause a load on the body part of the measurement target, The state value calculation unit calculates the load value based on the angle and duration from a reference position in the body part corresponding to the sensor attached to the measurement target based on the sensor data, The feature information is a representative time series pattern set representing each of the plurality of clusters based on an average time series pattern set obtained by averaging, for each element, a plurality of time series pattern sets included in each of the plurality of clusters, an analysis device.

2. The data acquisition unit acquires a plurality of sensor data from the measurement target, The clustering unit, For each of the plurality of attention intervals, generates a composite time series pattern by synthesizing time series patterns for each of the plurality of elements corresponding to the plurality of sensor data, Calculates the similarity between the generated plurality of composite time series patterns respectively, Performs the clustering of grouping attention intervals with close similarities as the same cluster based on the calculated plurality of similarities and generates the clustering result, the analysis device according to claim 1.

3. An operation type determination unit that determines the operation type of the measurement target based on the sensor data Is further provided, The clustering unit performs clustering using the state value for the plurality of attention intervals and the operation type and generates the clustering result, the analysis device according to claim 1.

4. The clustering unit, Classifies the plurality of attention intervals for each operation type, For each of the plurality of attention intervals for each operation type, generates a composite time series pattern by synthesizing time series patterns for each of the plurality of elements included in the sensor data, ​ ​ For each of the operation types, calculate the similarity between the plurality of generated synthetic time series patterns respectively. Based on the calculated plurality of similarities, perform the clustering for each of the operation types by regarding the target intervals with close similarities as the same cluster, and generate the clustering result. The analyzer according to claim 3.

5. The clustering unit generates the synthetic time series pattern by combining the time series patterns for each of the plurality of elements in the time direction. The analyzer according to claim 2 or claim 4.

6. The clustering unit generates the synthetic time series pattern by superimposing the time series patterns for each of the plurality of elements. The analyzer according to claim 2 or claim 4.

7. The clustering unit calculates the similarity using the DTW (Dynamic Time Warping) method. The analyzer according to any one of claims 2, 4, 5, and 6.

8. The predetermined criterion includes a first threshold value regarding a predetermined time length and a state value. The target interval setting unit extracts, from the time series data, an interval in which the state value exceeds the first threshold value over the predetermined time length, and sets the plurality of extracted intervals as the plurality of target intervals. The analyzer according to any one of claims 1 to 7.

9. The predetermined criterion includes a second threshold value smaller than the first threshold value. The target interval setting unit extracts, from the time series data, an interval in which the state value exceeds the second threshold value over the predetermined time length, and sets the plurality of extracted intervals as a plurality of candidate target intervals. The clustering unit performs clustering using the time series data of the state values regarding the plurality of candidate target intervals, and generates another clustering result. The analyzer according to claim 8.

10. The state value includes an LF / HF value representing the frequency component of the heartbeat variation as a ratio. The analyzer according to any one of claims 1 to 9.

11. A display control unit that displays display data based on the display information is further provided. The analyzer according to any one of claims 1 to 10.

12. The display data includes one or more videos related to each of the plurality of clusters and data of one or more regions of interest extracted from the time-series data corresponding to the one or more videos. The analysis apparatus according to claim 11.

13. The display data includes a representative video representing each of the plurality of clusters and data of a representative region of interest extracted from the time-series data corresponding to the representative video. The analysis apparatus according to claim 12.

14. The display data includes a plurality of videos related to at least one of the plurality of clusters. The analysis apparatus according to claim 11.

15. A computer acquires sensor data from a measurement target, calculates a state value based on the sensor data, sets a plurality of regions of interest in the time-series data based on the time-series data of the state value and a predetermined reference, performs clustering using the time-series data of the state value related to the plurality of regions of interest to generate a clustering result, generates display information including feature information of each of the plurality of clusters based on the clustering result and comprises the state value includes a load value numerically representing the degree of load in a working posture that can cause a load on a body part of the measurement target, calculating the state value calculates the load value based on an angle and a duration from a reference position in the body part corresponding to the sensor attached to the measurement target based on the sensor data, The feature information is a representative time-series pattern set representing each of the plurality of clusters based on an average time-series pattern set obtained by averaging a plurality of time-series pattern sets included in each of the plurality of clusters for each element.

16. A computer is caused to function as means for acquiring sensor data from a measurement target, means for calculating a state value based on the sensor data, means for setting a plurality of regions of interest in the time-series data based on the time-series data of the state value and a predetermined reference, means for performing clustering using the time-series data of the state value related to the plurality of regions of interest to generate a clustering result, means for generating display information including feature information of each of the plurality of clusters based on the clustering result and. The state value includes a load value that numerically represents the degree of load in a working posture that can cause a load on the body part to be measured. The means for calculating the state value calculates the load value based on the angle and duration from a reference position in the body part corresponding to the sensor attached to the measurement target, based on the sensor data. The feature information is an analysis program that is a representative time series pattern set representing each of the plurality of clusters, based on an average time series pattern set obtained by averaging, for each element, a plurality of time series pattern sets included in each of the plurality of clusters.

Images