Mowing Patterns Comparison: Analyzing the Mowing Behaviors of Elderly Adults on an Inclined Plane via a Motion Capture Device

Due to the declining population and the aging of the farmer population in the hilly and mountainous areas of Japan, it is necessary for the elderly to carry out the mowing work on the ridges and slopes, which is traditionally regarded as heavy labor as part of paddy farming. One of the most important causes of these accidents is an incorrect mowing posture; therefore, it is important and necessary to identify effective and safe working patterns during inclined plane mowing. In this paper, we designed and implemented a set of mowing experiments in a terraced field area in Hiroshima to collect information on the body motion of experienced elderly mowing workers via a high-precision motion capture device that supports the collection of information from 23 joints. According to an analysis that calculated the angles of the workers’ joints during mowing, we confirmed the characteristics of the mowing workers’ working patterns in three different situations (typical inclined plane mowing (TI), top-down mowing (TD) and bottom-up mowing (BU)). The comparative analysis indicates that the basic actions “c” (cutting) and “t” (throwing) are basically the same in terms of body posture for the situation TI, but for situation TD, the difference was observed with respect to the workers’ right ankle. Moreover, based on the comparation analysis for mowing action “c” (cutting), we confirmed that mowing workers should: keep their lower bodies as still as possible to keep balance for ensuring safety while working on inclined plane (TI); keep careful even if they are working on the flat ground (TD); and do not exert their utmost strength to mow, unless they are standing on the flat ground (BU). The findings of this work should be emphasized in the future development of mowing support systems and training programs for new mowing workers.


I. INTRODUCTION
In recent years, with the development of robot and AI (Artificial Intelligence) technologies, many studies have focused on systems such as automatic mowing machines [1]. However, the complex terrain of the hilly and mountainous areas in Japan requires that mowing work be done by manually operated machines [2]. Moreover, with the aging population increases in Japan, such mowing works are usually handed over to seniors [3].
The manually driven U-handle-type mowing machine is the most common device used to clear weeds on inclined The associate editor coordinating the review of this manuscript and approving it for publication was Zhong Wu . planes. Nevertheless, based on the research of the Japan Agricultural Work Safety Information Center, mowing is one of the tasks most prone to causing accidents in agriculture [4]. Among these, approximately 29.5% of accidents are caused by the unstable postures of the mowing workers [5]. According to past studies, incorrect postures cause a heavy burden on a worker's waist, shoulders, and hips, sometimes leading to falls or slips [6]. For elderly people, such accidents may result in death [6]. Studies on workers' dangerous postures and effective and safe working patterns during inclined plane mowing are therefore important and necessary.
However, few studies have focused on the analysis of body movements during mowing for elderly workers, especially when they work on an inclined plane covered with tall grass VOLUME 8, 2020 This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/ which hardly to observe. With the development of wearable sensor technologies, information on human body movement can be collected with high precision even in open environments [7]. Depending on the use of this type of motion capture device, the joints and related angles of mowing workers can be observed in a high precision, enabling the identification of low-risk mowing postures. Therefore, in this paper, we focused on the analysis of experienced senior workers' behaviors during inclined plane mowing and aimed to identify effective and safe working patterns in different situations that could reduce the risk of falls in different situations. The findings of this study may contribute to the development of future mowing worker training programs and mowing-related support systems.
The remainder of this paper is organized as follows: issues regarding mowing accidents in Japan and related research on human movement analysis in open environments are presented in Section II. In Section III, we introduce the proposed methodology, including the experimental design, dataset used, and variable measurements. Based on this, in Section IV, the results of the analysis and related discussion are provided. We offer the conclusions of this study and provide perspectives on future work in the field in Section V.

II. RELATED WORKS
According to studies on agricultural accidents in Japan, one of the most accident-prone agricultural activities is mowing [4]. One of the factors leading to accidents is age, because most of the workers involved in such accidents are elderly adults. According to the statistics on agricultural deaths from the Ministry of Agriculture, Forestry and Fisheries of Japan, workers over the age of 65 were responsible for 86.5% of agricultural accidents in 2018 [3].
Another important factor is worker posture. According to the research of the Ministry of Agriculture, Forestry and Fisheries in Japan, approximately 29.5% of these accidents were caused by unstable mowing postures [5]. Incorrect postures cause a heavy burden on the workers' waists, shoulders, and hips, sometimes leading to falls or slips.
Moreover, for mowing activities on inclined planes, according to research from the National Agricultural Research Center of Japan, because of the aging of the population, elderly adults are needed to perform mowing activities in rural areas. However, large lawnmowers are not useful on inclined planes, and mowing the grass with a manual lawnmower is inefficient [8].
Therefore, for mowing workers, especially older workers, knowledge about dangerous postures and effective and safe working patterns for inclined plane mowing is important and necessary. One possible method to develop safe and effective mowing postures is to analyze mowing workers' body movements via motion capture devices.
Given the characteristics of inclined plane mowing tasks, traditional optical-based motion capture devices, such as the devices used by Noiumkar et al. to analyze the golf swings of professional players [9], may not be useful because the supporting equipment cannot be deployed in the target terraced fields.
Therefore, some studies focused on human movement analysis in open environments. For example, to establish the feasibility of obtaining accurate outdoor kinematic data on competition-level road bicycles, Cockcroft et al. recoded and analyzed a set of motion data from ten male cyclists with their own bicycles on a stretch of road by using inertial motion capture systems [10]. However, the results showed that the data near the pedal and handlebar interfaces were unacceptable because of the magnetic interference of the system.
As motion capture devices continue to be improved, such temporary magnetic disturbances can already be fixed by interpolation. For example, a wearable acceleration sensor device such as an Xsens MVN Animate Pro (Xsens Technologies B.V.) can be used to collect human movement data with high precision even in a lab environment [7]. Moreover, unlike traditional motion capture devices (opticalbased or other acceleration sensor-based [11], [12]), which usually use a 17-joint model to represent the human body, the MVN device can collect data from 23 joints. These extra joints allow us to calculate more fall risk-related information, such as the angle of the inclined plane on which the worker stands.
Some studies have implemented the MVN system in the analysis of human movement analysis in open environments. For instance, Burget et al. compared healthy subjects and Parkinson's disease patients by requiring them to perform a hand coordination task with the MVN motion capture system. Their results indicated that different from clinical experience, the joint weights are almost evenly distributed along the arm in the PD group, but the proximal joint weights of the healthy subjects are notably larger than the distal weights [12]. However, few studies have focused on the analysis of body movements during mowing tasks on an inclined plane for elderly workers.
Therefore, to reduce the fall risk of elderly adults while mowing on an inclined plane, in this study, we analyzed the body movements of experienced subjects' mowing behaviors via the MVN system to identify potentially safe activity patterns and related techniques in different situations.

III. METHODOLOGY
To collect the target motion data and further test our proposed hypotheses, in this section, we will introduce the design of our implemented experiments and the related dataset and variable measurements used.

A. EXPERIMENTAL DESIGN
Based on an expansion of our previous research [14] and through cooperation with an agricultural corporation, a set of experiments for mowing were designed and implemented in a terraced field in Kouchi-cho, Higashihiroshima City, As shown in Fig. 1, the area selected for the experiment conforms to the following three characteristics: 1. the weeds in the area are too high to be cut efficiently with a regular lawnmower; 2. The grassy ridge area (inclined plane) is so long that workers have to stand on the inclined plane to mow the grass; 3. The two sides of the ridge area are flat, which means that some of the mowing can be done while standing in these flat areas. Since the main purpose of the experiment is to obtain motion data while mowing on an inclined plane, for this experiment, we asked the workers to mow the grass growing on the inclined plane with their usual style.
Moreover, it is worth noting that the inclined plane on which the experiment was conducted is so steep that an inexperienced ordinary person cannot even keep the balance to stand. As shown in Fig. 2, the inclination angle of the experimental inclined surface is larger than 45 degrees, and only experienced mowing workers can work on it safely. Such terrain is common not only in our experimental areas but also in the other hilly and mountainous areas of Japan. An Xsens MVN Animate Pro accelerometer-based motion capture device with a wearable 23-joint was used to assess the workers' mowing behaviors by having the system collect data including the 3D coordinate information of the subject's head, elbows, hands, waist, thighs, calves, and feet, with a frequency of 60 Hz. Moreover, since the motion capture device cannot record environmental information during the mowing action, with the consent of the participants, all experiments were recorded by video camera for verification purposes. Furthermore, although many types of lawnmowers exist, in this study, we focused on typical U-handle-type mowers, which are easy to equip and widely used in such ridge areas. The specific details are shown as Fig.3: Since the safety of the experiment was a major premise, only three professional mowing workers who are accustomed to mowing in this area participated in the experiment. All participants were over 60 years of age and healthy, had no problem in identifying right from wrong, and were right-handed. There was no time limit for our experiments, which were continued until the mowing workers finished their planned mowing tasks (almost 30 minutes). The details of participants are shown in Table 1. To improve the experimental results, related interviews were conducted to investigate the possible knacks or issues while mowing on an inclined plane. Finally, a set of questionnaires about the participants' physical data (body height, foot or shoe length, shoulder height, shoulder width, arm span, hip height, hip width, knee height, and ankle height) were administered to further improve the motion models.

B. DATA COLLECTION AND SELECTION
The experiments were carried out from September 17 to 19, 2019, during which the workers' mowing motion data were recorded. Finally, a total of 32,700 items of raw data were selected for analysis. The details of the data selection are described below.
As shown in Fig. 4, based on the video recording of the experiment and follow-up interviews, three types of situations were determined by the workers' standing posture and possible fall risk levels, which were named typical inclined plane mowing, top-down mowing and bottom-up mowing. These three types of mowing motion were analyzed and compared to identify their advantages and disadvantages while mowing on inclined planes. It should be noted that TI and TD were performed in the same orientation when cutting grass (from left to right), but BU was performed in the opposite orientation. It is well known that to perform a mowing task, two basic actions are required: lifting and cutting. Since there is no resistance from the grass during the lifting motion, we only focused on the cutting motion aspect. The remaining motion data were excluded from this study.

C. VARIABLE MEASUREMENTS
Based on the manual of the Xsens MVN Animate Pro (the accelerometer-based motion capture device), which can record the joints' 3D coordinates every 4 milliseconds (ms) [7], the 3D coordinate data from a total of 23 types of joint can be exported as xml files. In contrast to the traditional 17-joint models, such as those that produced the Human3.6M dataset [15], more joints, such as the subjects' toes and spinal segments L5, L3, T12, and T8, can be exported, which means that more information can be calculated based on the extra joint data, such as the subjects' ankle and waist bending angles.
Based on past work and the characteristics of our study, three types of metrics were used to describe the mowing workers' body motions, which can be calculated by the 3D coordinate data of their joints: the joint angles, moving distance and speed, and moving distance along the Z-axis.
As shown in Fig. 5, the joints of a subject's arms, waist, legs, and feet are selected to represent the mowing posture. Therefore, based on these joints, a total of six bending angles were calculated in this study, including those of the wrist, elbow, waist, knee, and ankle. Next, since data from the subjects' feet and toes are also collected by this system, we used three of the points (both sets of toes and the weight-bearing foot, not on the same level but with a constant angle) to calculate the angle of the inclined plane, which can be used to represent the risk level of subjects to a certain extent. Moreover, based on the coordinates of the subject's hands in each frame (4 ms), their moving distance can be calculated to represent the strength of the subject's mowing action. The details of the calculations are provided below.

1) JOINT ANGLES
As shown above, based on the angle formula [16], the joint coordinates can be used to calculate its bending angle. Taking the wrist angle (θ wrist ) as an example, assuming the subject's 3D coordinates are E (X e , Y e , Z e ) for the elbow, Then, the angle θ wrist will be: Following the same calculation, five angles can be obtained, θ waist , θ elbow , θ wrist , θ knee and θ ankle , to represent the subject's posture. Excluding the angle of the waist, data for the other angles were collected in pairs (e.g., right and left hands).
Moreover, since the calculated angles linear change throughout the mowing action, the range of the angles was determined to characterize them. For example, for the bending angle of the worker's right elbow, we calculated Rng Relbow for each mowing action. The exception was the waist angle, whose movement is not linear; thus, the standard deviation Std waist was used to characterize the range of this angle.

2) ANGLE BETWEEN INCLINED PLANE AND HORIZONTAL LINE
Since information on the subjects' toes and feet can be obtained in our used system, two points corresponding to the toes of the subject's two feet and one corresponding to the foot that bore the weight of the entire body were used to calculate the angle between the inclined plane and a horizontal line (S&H angle).
Assume the coordinates of the subject's left toes are L(X l , Y l , Z l ), those of the right toes are R(X r , Y r , Z r ), and those of the weight-bearing foot are F (X f , Y f , Z f ). First, we need to calculate the normal vectors − → FL and − → FR to represent the inclined plane: Therefore, Thus, According to the law of sines, the angle between the inclined plane and the horizontal line θ sh will be: In this study, we use the standard deviation of a subject's S&H angle value Std sh to measure the shaking level during mowing.

3) MOVING DISTANCE OF THE HANDS
In analyzing mowing tasks on inclined planes, the most important aspect is the mowing workers' hands, which come into direct contact with the mowing device. Generally, the mowing worker's hands did not move in a straight line while mowing. We defined and calculated the moving distance of the hands as follows: Taking point i with a duration of 4 ms as an example, the coordinates of the hands before i and after i are assumed to be Be (X Be , Y Be , Z Be ) and Af (X Af , Y Af , Z Af ), respectively. The moving distance D i of the hand will be: Therefore, if the target period of one mowing action consists of n points, each with a duration of 4 ms, the moving distance of the hands will be: Therefore, for both workers' hands, we can obtain separate results of the moving distance in the 3D coordinate system, D right and D left ,to represent the strength of the mowing actions.

IV. ANALYSIS
As mentioned in section III, in this study, we classified the workers' mowing behaviors into three categories according to their position relative to the inclined plane: typical inclined plane mowing (TI), top-down mowing (TD) and bottom-up mowing (BU). In this section, we aimed to analysis these three situations respectively. Firstly, based on the related data analysis and interview results, identify mowing workers' basic mowing actions and their character and confirm the similarities and differences between these basic actions via Mann-Whitney U-test. Finally, the related variable measurements will be compared among these three situations by using the Kruskal-Wallis test. The relevant analyses were performed on SPSS Statistics 25.

A. TYPICAL INCLINED PLANE MOWING
In typical inclined plane mowing, the worker stands on the inclined plane to mow the grass while maintaining balance. This action was typically performed when the ridge area was too broad to allow the workers to mow on the top or bottom of the ridge.
Based on the related interview results and the comparison between the experimental video recording and the MVN-based 3D model recording, during inclined plane mowing, three basic actions were identified according to their goal: cutting (c), throwing away grass (t) and moving (m). The cutting action is the strongest, but sometimes just one cutting action could not achieve the desired mowing effect. Then, since there were still uncut grasses at the lower end of the mowing area, the mowing workers had to throw away the grass down the inclined plane through a few actions. We recorded the order of the workers' actions shown in Table 2. According to the results, several characteristics can be observed. First, when mowing was completed for one area, the worker moved to the next area; during TI, no mowing actions were done while ''moving''. It was then possible to identify mowing action patterns on the inclined plane. In detail, if the ''moving'' action is used to indicate the end of a set of actions, whole actions (''c'' and ''t'') can be classified into many groups of similar units of actions: action ''c'' always occurs before action ''t'', before moving to the next area, and workers usually performed action ''c'' 1∼4 times and action ''t'' 1∼5 times to finish mowing an area.
However, although the purpose of actions ''c'' and ''t'' is different, the difference in body motion between these two types of actions was not clear, and a detailed analysis was required. Therefore, the Mann-Whitney U test was used to identify the difference between actions ''c'' and ''t'' according to the measurements we calculated.
As shown in Table 3, the results indicate that except for the measurements D right (p < 0.05; Estimate: −0.100; 95% CI: −0.148, −0.039) and D left (p < 0.05; Estimate: −0.059; 95% CI: −0.104, −0.014), there is no difference between actions c and t at a significance level of 5%. The experimental results show that although the mowing actions ''c'' and ''t'' look different in the video recording, the results of comparative analysis indicate that they are basically the same in terms of body posture. However, motion t required the workers to use more power (a higher value of the hand-moving distance) to throw the grass away.

B. TOP-DOWN MOWING
In top-down mowing, the worker stands at the top of the inclined plane to mow the grass while maintaining balance. This action typically occurred when a flat area was found at the top of ridge area from which the mowing workers could stand to do their work. However, based on the interview and related video recording, to work more efficiently, instead of standing on the flat area the entire time, the workers sometimes stuck their feet out onto the inclined area to do their work.
Similar to the TI situation, in this situation, three basic actions were observed, and we identified them by their goal: cutting (c), throwing away grass (t) and moving (m). Workers still had to throw away the grass down the inclined plane. We recorded the order of the workers' actions, which is shown in Table 5.
According to the results, some characteristics can be observed. First, similar to TI, in TD, workers also performed ''moving'' actions, during which no mowing actions occurred. A similar mowing action pattern can also be observed in situation TD; action ''c'' always occurs before action ''t''. Finally, different from situation TI, in situation TD, workers used action ''c'' 2∼7 times and action ''t'' 1∼2 times to finish mowing an area. The number of ''t'' actions in situation TD is significantly smaller than that in situation TI.
Similar to TI, although the purpose of actions ''c'' and ''t'' is different, the difference in body motion between these two types of actions is not clear, and a detailed analysis was required. Therefore, to identify the difference between these two types of actions, the Mann-Whitney U test was applied.
As shown in Table 4, the results indicate that as in situation TI, measurements D right ( The experimental results show that similar to situation TI, the moving distance of the hands is also different between mowing actions c and t in situation TD. However, a difference in body posture was also observed with respect to the   workers' right ankles and left knees. As shown in Fig. 6, the position of the grasses is too low in this situation, so the workers need to maintain a semi squatting posture to throw the weeds with great force.

C. BOTTOM-UP MOWING
The last situation we discuss is bottom-up mowing. In this situation, the worker stands at the bottom of the inclined plane to mow. Compared to the other two situations, the BU situation may be the safest way to mow; however, in most cases, the weeds grow too high on the inclined plane to be cut with this action.
Different from the TI and TU situations, only two basic actions were observed in the BU situation, which were identified by their goals as standing cutting (c) and moving cutting (m). During BU, the workers did not need to throw away the grass because there is nothing to stop the grass from falling freely. We recorded the order of the workers' actions, which is shown in Table 6. Notably, although we used the same tags ''c'' and ''m'' as with TI and TD, it does not mean that the actions were similar to those of the previous two situations. VOLUME 8, 2020  According to the results, we can identify that, unique for the BU situation, the workers mowed the grass while performing the ''moving'' action. However, no particular pattern can be observed among the ''c'' and ''m'' actions. Each action may be seen as independent mowing patterns. Nevertheless, a related analysis was required to identify the possible difference. Mann-Whitney U tests were applied to identify the differences between these actions.
The experimental results indicate that for almost all of the measurements, a significant difference was not observed between motion t and m, except for the right knee and ankle. Moreover, the value of the two metrics was greater for action m than for action ''c''.
As shown in Fig. 7, the reason for the difference may be that the workers use their left leg for weight-bearing while their right leg moves with the cutting action.
Since the workers in this situation did not worry about falling, the collected motion data were used as a reference for the other two situations.

D. COMPARISON OF THE THREE SITUATIONS
As discussed above, the three types of mowing situations have their own characteristics in terms of actions, such as action ''t'' in TI & TD and ''m'' in BU. However, they all share the normal cutting action ''c'' in common as the most important action for mowing. Therefore, in this section, we aimed to compare the three types of mowing action ''c'' for all calculated measurements.
Notably, before performing the comparisons, we adjusted the data from BU to make the comparisons more meaningful (the left and right measurement data were swapped) because during BU, the workers operate using the opposite orientation from that of the other two situations.
The Kruskal-Wallis test was used to compare the measurements among the three situations except for Std sh . The Mann-Whitney U-test was used to compare Std sh between situations TI and TU because this measurement does not mean anything for situation BU.  Tables 8 and 9, for the comparison of the joint angle ranges among the three situations, the results indicate that except for the workers' left knees and ankles, significant differences were not identified for the remaining measurements.

As shown in
To be specific, according to the related paired comparison analysis, situation TD was not different from situation BU, as both have smaller angle ranges than situation TI. A possible explanation is that both TD and BU involve mowing on flat ground, while the workers need to keep their left knees and ankles within a small range for balance when mowing on the inclined plane (situation TI).
On the other hand, the comparison of Std sh (p > 0.05; Estimate: 0.036; 95% CI: −0.231, 0.431) between TI and TU showed that for these two situations, the workers had similar shaking levels of the angle of the plane which they are standing on.
Moreover, different from the other two situations, the results also show that situation BU produced the largest value of Std waist , which may indicate that the workers exerted their utmost strength to mow in this situation (that is, a higher level of waist shaking was employed). The results of the comparison of the right-hand moving distances also confirm this inference.

V. DISCUSSION
Based on an experimental analysis of three situations, typical slope mowing (TI), top-down mowing (TD), and bottomup mowing (BU), with a comparison of these situations, the following inferences were confirmed. The discussion below is based on the assumption that experienced mowing workers exhibit safer mowing behaviors than novice mowing workers.
The analysis results of three situations TI, TD and BU indicate that work patterns based on basic actions depend on the position of the worker on the inclined plane. However, the results of comparison analysis among these three situations indicate that except for the workers' left knees and ankles, significant differences were not identified for the remaining measurements.
Therefore, considering that the motion data we collected were from professional mowing workers, we can infer that when mowing grass from right to left, among all parts of the body related to posture, the most in need of attention is the left leg (that is, the weight-bearing leg; opposite results were observed when cutting from left to right).
Which means that for the mowing posture, mowing workers should keep their lower bodies (especially the load-bearing feet) as still as possible to keep balance and further reduce the possibility of accidents while mowing on inclined plane (TI). Second, for the shaking level while mowing, the paired comparison analysis results observed similar results between TI and TU, which means that mowing workers should keep careful because of the continual posture changing even if they were standing on the flat ground above the inclined plane (TD).
Moreover, for the mowing actions' intensity and range, paired comparison analysis results showed that the BU is observed as the biggest, which means we can infer that the mowing workers should not be allowed to exert their utmost strength to mow unless they are standing on the flat ground (BU).
As an application scenario, when a new mowing worker gets ready to mow in the rice terrace area, before the mowing actions (including the situations of after each moving action), he/she needs to make sure that the weight-bearing foot is firmly planted to keep the body balance. While mowing, the new worker also should not relax his/her vigilance and mow the grass in accordance with the established patterns. For the reason that even though he/she is working on the top of the inclined plane, it is also needed to change the postures continually which may be dangerous. Moreover, unless mowing on the flat ground, the new worker needs to use his/her strength carefully to avoid upsetting the balance.
On the other hand, for the further development of a mowing support system, it is necessary to design a module that detects the stability of the weight-bearing foot. It is also expected to have the ability to warn users that they should not continue mowing unless they pass the sensors' stability test. Besides, another module which can detect the angle between the feet is also needed to monitor and predict users' possible dangerous level by comparing the data collected in normal stable postures.
What is noteworthy is that the results we presented above have been reconfirmed via the experienced mowing workers' interviews while the return visit to Hiroshima in 2020.9.

VI. CONCLUSION
This paper focused on an analysis of elderly workers' body movements during inclined plane mowing. Based on the design and implementation of a set of mowing experiments, we collected a total of 32,700 items of motion data from elderly experienced mowing workers to calculate the angles of the workers' joints and the inclined plane for analysis via the high-precision motion capture device Xsens MVN Animate Pro.
Compared to other wearable sensors-based studies on the human behaviors' analysis which more focus on the measurement accuracy [17] and human action classification [18], our study focuses more on the comparation of different actions and the identification of the cause-and-effect relationship between human posture and falling risky.
The results of the analysis indicate that according to the position where workers stand relative to the inclined plane, different characteristics (different basic actions) of working patterns are observed. To be specific, the analysis indicates that the basic actions ''c'' (cutting) and ''t'' (throwing) are basically the same in terms of body posture for the situation TI, but for situation TD, the difference was observed with respect to the workers' right ankle.
Moreover, based on the comparison among all three situations (TI, TD and BU) for mowing action ''c'' (cutting), the results indicate that to ensure safety, mowing workers should keep their lower bodies (especially the load-bearing feet) as still as possible to keep balance while working on the inclined plane (TI); need to keep careful even if working on the flat ground above the inclined plane (TD), because the shaking level of standing plane is same to TI; do not exert their utmost strength to mow unless stand on the flat ground (BU). The findings here should be considered when developing future mowing worker training programs and mowing support systems.
This study is currently in the process of data accumulation. The limitations of this study are that because of an inexperienced ordinary person cannot even keep the balance to stand on the experimental inclined plane, to ensure the safety of the experiment, we only invited three experienced mowing workers to participate our experiments. Moreover, this study did not consider the influence of the working environment except the inclined plane, such as objects in the grass and the subjects' eye movements [19].
To contribute the future mowing assistance system development and education curriculum, in future work, we will improve our experimental design, invite more experienced mowing workers to participate in the experiments. Moreover, we also plan to use of depth camera and eye-tracking device sensors to enrich our data. In addition, we will consider using machine learning to predict mowing workers' high-risk behaviors.
BO WU (Member, IEEE) received the Ph.D. degree in human sciences from Waseda University, Tokyo, Japan, in 2015.
He is currently an Assistant Professor with the Department of Human Informatics and Cognitive Sciences, Waseda University. He is also a Research Follow with the DS Laboratory, Kansai University. His current research interests include human motion capture, eye-movement analysis, machine learning and big data analysis, human-centered application system development, the Internet of Things (IoT), information and computer science, and social and human informatics. He is a member of the IEEE CS and the Information Processing Society of Japan.
YUAN WU received the Ph.D. degree in human sciences from Waseda University, Tokyo, Japan, in 2016. She is currently an Assistant Professor with the Department of Human Behavior and Environment Sciences, Faculty of Human Sciences, Waseda University, Saitama, Japan. Her research interests include endogenous development in less favored areas, especially studying the agricultural resources management in the hilly and mountainous areas on the community basis of Japan, smart farming, and potential of information and communication technologies (ICT) into the maintenance and management of agricultural resources.
YOKO AOKI received the master's degree in pedagogy from The University of Tokyo, Japan, in 2006.
She is currently a Research Associate with the Department of Human Behavior and Environmental Sciences, Waseda University. Her research interests include tool using behaviors at meal-settings in infancy and the development of bimanual manipulation in natural context and learning process through the interaction between human and the environment. She is a member of the Japanese Society of Ecological Psychology, the Japanese Society of Developmental Psychology, the Japanese Association of Qualitative Psychology, and the Japanese Association for Medical and Psychological Study of Infancy.
SHOJI NISHIMURA (Member, IEEE) received the bachelor's degree in mathematics and the M.Sc. degree in applied physics from Waseda University, Tokyo, Japan, and the Ph.D. degree in human sciences from Osaka University. He joined the Advance Research Center, INES Corporation, as a Senior Researcher in 1991. He joined the School of Human Sciences, Waseda University, as an Assistant Professor in 1997. He is currently a Professor with the Faculty of Human Sciences, Waseda University. His current research interests include educational technology, especially education and the Internet. He is a member of the Japan Society for Educational Technology, the Japanese Society for Information and Systems in Education, and the Information Processing Society of Japan. VOLUME 8, 2020