In recent years, fruit theft has become a serious social problem in Japan, where conventional surveillance measures are limited due to the unique challenges in agricultural areas, such as limited power supply and large areas. To address this, previous studies have developed battery-powered, microphone-based theft detection systems that operate on microcontrollers; however, achieving high classification accuracy under severe resource constraints remains a challenge. This paper presents a design and training methodology for a compact, high-accuracy sound classification model for fruit theft prevention. Our approach combines Depthwise Separable Convolution (DSC) for efficient feature extraction with Ensemble Distillation (EnD), which transfers knowledge from two teacher models with distinct feature extraction approaches. Experiments on a three-class task (environmental sounds, speech, and footsteps) demonstrated that the proposed model achieved an F1-score of 88.61%, an improvement of 12.00 percentage points over the baseline. The inference memory usage was limited to 49 kB, and including the entire system for processes other than deep learning inference, the total memory usage remained within 160 kB, well below the 256 kB RAM capacity of the target hardware. Moreover, the model achieved an inference time of approximately 0.4519 seconds per 1-second audio segment on a Seeeduino XIAO nRF52840 micro-controller. These results confirm the practical feasibility of the proposed system as a robust AI-based surveillance solution under strict resource constraints for real-world orchard applications.

Currently, skilled farmers judge the ripeness of the Shine Muscat grape variety by looking at the color on the surface of the grapes. However, the color of Shine Muscat grapes does not change much as they grow, and there are individual differences in the way the color is perceived. Furthermore, the same color can look very different depending on the exposure to sunlight and shadows. Therefore, there is a need for a system that can quantitatively determine the color of Shine Muscat grapes to pass on the harvesting techniques of experienced farmers to amateurs and inexperienced farmers. This research aims to improve the accuracy of the color estimation of Shine Muscat grapes using deep learning. We propose a method to estimate the color of individual grapes using a color estimation model with a self-attention mechanism, from which the color of the whole bunch is estimated. A Vision Transformer model with a self-attention mechanism was found to improve the color estimation accuracy to $96.9 \%$. Furthermore, by eliminating outliers using the interquartile range, a color estimation accuracy of $97.2 \%$ could be achieved, demonstrating the effectiveness of the new color estimation model.

There is frozen shoulder that a patient feels pain around the shoulder if the shoulder is moved in everyday life. It is necessary for the patient with the frozen shoulder to receive the rehabilitation for improvement of the QOL in the hospital, and it is important that the load of the physical therapist is reduced. To solve them, various machines and robots of the rehabilitation for the frozen shoulder are studied and developed. It is not suitable to repeat predefined motions to trace the general rehabilitation motion, since the motion of the shoulder in the rehabilitation should be adapted to the condition of each patient. Therefore, we are developing the rehabilitation robot for the frozen shoulder based on the direct teaching and playback motion. This paper describes the development of apparatus adapting each motion in the rehabilitation in the hospital normally. First, the motions in the rehabilitation are analyzed, the mechanism of the apparatus is considered based on the analysis. After that, the apparatus with five actuators is developed practically, and the direct teaching and playback motion are investigated by the participant that wears the apparatus. As a result, the realization of the apparatus is verified.

In Japanese fruit-growing regions, fruit theft is a significant issue. Traditional anti-theft methods are fraught with various shortcomings and are often ineffective. To address this, we have developed a new approach for preventing fruit theft: a device that detects suspicious sounds by integrating a sound sensor with a compact, low-power microcontroller. This paper presents a suspicious sound detection system designed for a fruit theft alert device. It details the development of a deep learning model for detecting suspicious sounds, covering aspects from data collection to model training and evaluation. Despite operating on a microcontroller with limited memory and computational power, the system under development has achieved a classification accuracy of up to 78.5% in F1-score for footsteps, speech, and other environmental sounds.

Abstract The color of a bunch of grapes is a very important factor when determining the appropriate time for harvesting. However, judging whether the color of the bunch is appropriate for harvesting requires experience and the result can vary by individuals. In this paper, we describe a system to support grape harvesting based on color estimation using deep learning. To estimate the color of a bunch of grapes, bunch detection, grain detection, removal of pest grains, and color estimation are required, for which deep learning-based approaches are adopted. In this study, YOLOv5, an object detection model that considers both accuracy and processing speed, is adopted for bunch detection and grain detection. For the detection of diseased grains, an autoencoder-based anomaly detection model is also employed. Since color is strongly affected by brightness, a color estimation model that is less affected by this factor is required. Accordingly, we propose multitask learning that uses metric learning. The color estimation model in this study is based on AlexNet. Metric learning was applied to train this model. Brightness is an important factor affecting the perception of color. In a practical experiment using actual grapes, we empirically selected the best three image channels from RGB and CIELAB (L*a*b*) color spaces and we found that the color estimation accuracy of the proposed multi-task model, the combination with “L” channel from L*a*b color space and “GB” from RGB color space for the grape image (represented as “LGB” color space), was 72.1%, compared to 21.1% for the model which used the normal RGB image. In addition, it was found that the proposed system was able to determine the suitability of grapes for harvesting with an accuracy of 81.6%, demonstrating the effectiveness of the proposed system.

The color of grape bunches is a significant factor when harvesting grapes at the appropriate timing. Judging the suitable color for shipment requires experience and varies from one person to another. We herein describe a support system for grape harvesting based on color estimation. To estimate the color of a bunch of grapes, bunch detection, grain detection, removal of diseased grains, and color estimation should be performed. Models based on deep learning are employed for this series of processes. Since color is strongly affected by sunlight, we propose a multitask model that considers sunlight exposure to achieve a robust color estimation model that exhibits decreased sensitivity to sunlight. Our results show that the color estimation accuracy of the model is 76% when sunlight exposure is not considered and 81% when sunlight exposure is considered. In addition, we performed a practical field test of the developed harvest support system in an actual grape field. The results show that our support system can determine the appropriateness of grape harvest with an accuracy of 90%, demonstrating the effectiveness of the system.

Inflorescence trimming is a crucial process to produce high-quality table grapes. It can eliminate nutrient competition in a bunch and makes it less vulnerable to disease development. After trimming, the remaining part of the inflorescence should have a target length decided by the grape variety. This is challenging for novice farmers because of the time constraint. The farmer needs to finish trimming the inflorescence before the berries develop. This paper proposes a novel end-to-end inflorescence length measurement method for supporting a trimming process with augmented reality technology. The proposed technique makes use of the state-of-the-art deep neural network model for detecting the inflorescence area, as well as the scissors from the images captured with a camera installed on an optical see-through head-mounted display. A new algorithm is designed to estimate the length of the remaining inflorescence with the screw of the scissors loop as the calibrator. The estimated length is then visualized on the head-mounted display to support the farmer in performing the trimming correctly and efficiently. The experiment, conducted with real inflorescence trimming tasks, shows that the mean absolute error of the length estimation is only 0.19 cm, which is small enough for use in real applications.

Augmented reality (AR) is a technology that expands information by superimposing digital information, such as virtual objects, on the real world using smartphones, smart glasses, and head-mounted displays. It is used in a variety of situations. In this paper, we propose a system that allows vine farmers to investigate effectively the vegetation condition of trellising-style vineyards using a head mounted display and AR technology. The experiment results show that by using a hybrid navigation approach include showing the whole vineyard in a small window and showing the details only, when necessary, the proposed system enable the farmers to move to a location with concern accurately.

In past research, we have developed a telescopic rod system that expands and contracts inside the human body that can be used for spinal fusion. The system can be expanded and contracted as the wearer grows and changes in symptoms by applying ultrasonic vibration. The telescopic rod system consists of instruments used in actual spinal fusion, special metal rods, and metal nuts. The tightening torque of the locking cap that secures the metal rod to the spine changes the movement of the rod system. In this paper, we show the verification of the effect of tightening torque on the movement of the rod. Verification was performed in water using an ultrasonic probe. From the experimental results, the effect of tightening torque on the outside was quantitatively shown, and the results effective for future development were shown.

To treat scoliosis, a surgery is conducted where a special metal rod is attached inside the human body. Surgery requiring an incision is performed regularly; thus, patients face the risk of infection and the discomfort of surgery multiple times. To solve this problem, a rod system that expands and contracts inside the human body has been developed. However, it is not a structure suitable for being implanted in the human body. The purpose of this research is to develop a new rod system that solves this problem. In this paper, the structure of the new rod suitable for being implanted was shown, and the analysis results using ANSYS were shown. In addition, the results of a verification experiment using a laser displacement meter were shown, indicating the superiority of the new rod.

Rheumatoid arthritis is a disease of the joint that are destroyed, it is difficult for a patient with serious condition to live his or her everyday life. It is important to evaluate the condition of rheumatoid arthritis in order to give the suitable treatment. However, the evaluation task takes time and is necessary to experience of the doctor. Therefore, it is desirable to develop the automatic evaluation system. Our objective goal is to develop the automatic evaluation system that can be updated using the revised data obtained by the doctor. It is clear that the evaluation system of the doctor consists of the recognition system and the classification system. This paper proposes the recognition of the joint in the X-ray photograph using the machine learning. To realize the system, we separate the recognition system into four procedure; convert procedure, training procedure, validation procedure, and feedback procedure. And the effectiveness of the proposed system is investigated using the real X-ray photographs of the patients with rheumatoid arthritis. As a result, it is clear that a lot of correct data are necessary to improve the accuracy. Therefore, it is clear that the it is more effective to improve the accuracy, if the revised data obtained by the doctor are feedbacked to the training system.

A pincer nail is a disease of the form of the nail, it is difficult for a patient with a serious pincer nail to walk. There are a lot of opinions in the cause of the pincer nail, however the cause is not clear. Some of the medical doctor guess that one of the cause is the gait motion from the implicit knowledge. The implicit knowledge is useful and important, however it is not expressed by the sentence and figure. This paper proposes the method for prediction of the attention point in gait motion using the factor analysis. Three attention points are obtained by the result of factor analysis. It is clear that two attention point among three are corresponding with the previous studies. We recheck the movie, since the rest of the attention point is considered. The rest of the attention point is the new unique gait motion that we have not noticed. As a result, the result of the factor analysis is useful to extract and guess the factor of the unknown factor.

The accuracy of the inspection of the capsule endoscope is increased, if it can be manipulated by the operator with wireless communication. To develop it, the feasibility is important, however there are various studies. We consider the behavior of the endoscope in the broken-down. Therefore, this paper describes the manipulatable capsule endoscope that can behaves as the normal endoscope even if it is broken in the body of the patient, and that is not modified from the normal endoscope drastically. The fin for the maneuverability is knit using the biodegradable string for the surgical operation which is dissolved in the body, since the various shape can be realized. The safety is guaranteed, since the fin is dissolved in case that it comes off in the body. And, the small motor is employed as the actuator for the movement of the fin in the fundamental experiment to prevent from changing the shape of the capsule endoscope. The proposed endoscpoe can behave as the normal capsule endoscope, since the shape is similar to the normal capsule endoscope. Using it, the feasibility of the proposed endoscope is confirmed by the fundamental experiments.

One of the treatment for the scoliosis is the surgery for attaching a metal rod to spines. It is necessary to conduct the surgery again if the patient grows up. We have been developing the new mechanism rod that enables the length of the gap of the spina to be controlled using ultrasonic vibration. It is important to design the horn to transmit the ultrasonic vibration to the rod effectively. This paper describes the design guideline of the horn for adapting the developed controlling rod system. First, the requirement of the horn is arranged, and the guideline is shown. Next the vibration modes are considered if the shape of the horn is not satisfied for the design guideline. Using the consideration, we confirm the importance of the design guideline. Finally, the actual horn is developed along the proposed design guideline, the stable vibration is verified.

All generation person should enjoy playing sports to keep and improve health. It is important to instruct a beginner in the techniques of sports to enjoy a sport. However, instruction for the beginner is challenging because of the difficulty of evaluation of the motion. In this paper, the classification for the evaluation of the motion in the sport is addressed using the recurrent-based neural network that is one of the deep learning. Moreover, this paper deals with the swing motion of tennis, since the swing motion of the tennis is essential, and the beginner is often instructed in the swing motion at first. First, a developed measurement system for human motion is described. Next, the recurrent-based neural network to classify the swing motion is shown. Final, the classification results are discussed. The individual swing motion can be classified using deep learning framework. However, it is clear that the swing motion of the experienced player is not always the same. Therefore, we confirm that individual instruction is important to improve the motion of the sport.

The measurement of walking is important for evaluating the health of elderly people. This evaluation is performed by experts in gait assessment (such as doctors or physiotherapists), who vary in terms of skill. To guarantee evaluative quality, engineering of gait-measurement solutions is desired. This paper describes a method of quantifying tacit physiotherapist knowledge of walk rehabilitation by tracking and analysing their eyes. We constructed the eye-tracking system to extract the points of gaze of multiple physiotherapists. The result shows that proficient physiotherapists exhibit special eye movements, which we defined as `rhythmical'. We evaluated the proficiencies of all subjects based on their rhythmical-eye movement. It is clear that proficiency does not increase linearly with rhythmical-eye movement, but instead exhibits a staircase-like behaviour. Consequently, we elucidated that the `rhythmical-eye movement' is one part of tacit knowledge of the proficient physiotherapists. And it is useful to establish the walking measurement closer to the proficient physiotherapists.

It is important to automate a heavy and hard task for workers. One of the hard tasks is that the worker picks up the bed sheet after washing, spreads it and enters it to ironing machine. In Japan, washing machine and ironing machine are working in the factory. However, the task such that the bed sheet is entered to ironing machine is not automated yet. We have developed handling machine to spread the bed sheet. However, there are some problems in successful rate and time cost. This paper describes the new handling machine to solve the problems. The new handling machine include the new algorithm and new robot arm is developed. And experiments are performed using real bed sheets. As a result, it is confirmed that the successful rate is increased, and time cost is decreased.

Increasing the export of fruit is vital for the Japanese economy. In this study, we focused on the peach trade as it is the third most transported fruit in Japan. The peach is a very sensitive fruit; therefore, proper packaging is important. We proposed a special packing sheet that we named the “concavo-convex sheet” developed through a simple mechanism. However, there were some problems in creating this sheet. This study describes the new mechanism for creating the concavo-convex sheet, and a real machine using the installed mechanism was developed. As a result, it was confirmed that the sheet can be realized by the new machine.

In this paper, a finger force distribution measurement system for hand dexterity is developed, newly. This system consists of a measurement device, wireless communication units and a display unit. Measurement data are expressed on the display of the PC in real time and are stored. The other almost devices for finger force and hand force measure the maximum force. Our developed device measure the each finger force distribution, when a cylinder form thing such as a cup or a pet bottle is brought up. Using the measurement system, each finger force of some subjects that is healthy and young are measured to confirm the function of the device. The validity of the device is verified by the experiments. It is clear that influence of little finger is large and the position of thumb is important in grab motion.

An important goal for the Japanese economy involves increasing the export of fruits. Especially, the export of peaches to Taiwan is an important part of trade for Japan. However, the peach fruit moth poses a serious problem. The moth is not native to Taiwan. Therefore, the moths that are exported with peaches may cause ecological damage in Taiwan. We are in the process of developing a peach fruit inspection system involving the use of X-rays. A handling unit that is a part of the inspection system is required to softly hold the peach without damaging it. This study describes a concavo-convex sheet developed by two nonwoven fabric clothes that is adapted for the inspection system. There are two important factors involved for incorporating the sheet in the inspection system, namely, the friction coefficient of the sheet and damage protection for the peaches. These factors are investigated by performing experiments that use several real, fresh peaches. The results confirm that the concavo-convex sheet satisfies the important properties necessary to employ the system for peach fruit inspection.

This paper describes an unique sheet developed from non-woven fabric that can be used to reduce transportation shock experienced by fruits that are easily damaged, such as peaches or strawberries, when transporting to foreign countries where there is a significant demand for these fruits. The proposed sheet comprises two non-woven fabric cloths suitable for packing these types of fruit. The non-woven cloth has useful characteristics such as air permeability, sealing properties, and X-ray transmission properties that make it suitable for transporting easily damaged fruit. In this paper, the application of the proposed sheet is applied for the packing of the fruits, however, the sheet that has the air permeability and shock resistance property is useful for the assistive robot and human interaction robot. And, the sheet is made manually, since the method to produce it is not constructed. The task of the human is reduced, if the mechanism for making the sheet can be realized. This paper describes that the properties of the sheet are introduced, and that the performance is investigated by experimental methods. Finally, the mechanism for producing the sheet is described and investigated using an experimental prototype.

Patients with Extension Lag cannot extend their knee straightly single-handedly. However, they can extend their knee without pain, if the other (a physical therapist (PT)) stretch the knee. The recovery of the Extension Lag is not necessary surgery, and is often enough only to perform rehabilitation. However, the rehabilitation of the Extension Lag is hard work for the PTs. This paper proposes the control algorithm for the automatic rehabilitation of the Extension Lag by adapting the Knee Assistive Instruments for Rehabilitation (KAI-R) that has been developed for the walking rehabilitation after the surgery of the Total Knee Arthroplasty (TKA) in order to reduce the load of the PT. The safety and the effectiveness of the robot are necessary to be verified, because we are going to adopt the robot to a clinical trial. First, the validity of the motion of the robot that is installed the control algorithm is confirmed by the high speed movie camera. Secondly, the effectiveness is investigated by using surface myoelectric potential. Thirdly, the motion of the robot is confirmed by the doctor and the PTs. It is clear that the robot for the rehabilitation of the Extension Lag can be adopted to a clinical trial.

It is important to evaluate the degree of functional recovery on a walk in rehabilitation. We have proposed two kinds of evaluation methods. One is the Gravity Center Fluctuation (GCF) obtained by the developed special shoes. The other is five element score method using the high speed camera, and is reliable to evaluate the functional recovery. In this paper, we consider the relation between the GCF and the score by the Sell-Organizing Map (SOM). As a consequence, we show the condition of the gait motion is able to be guessed by the SOM.

This paper describes a model of using a digital hand, which mimics human hands, operates dynamically in real time operation. Focusing on real time operation, we consider a model structure of digital hand with reduced DoFs (degree of freedoms) as an approximated model, where the reduction is based on the anatomical and medical hand analysis. There some problems because of the approximated model. To overcome the problems, some techniques are implemented into the model. We examine how the model is able to mimic the movement of human hands.

In this research, a classification method for human touching behaviors to haptic therapy robot is proposed. It is difficult to apply explicit discrete representation for human touch behaviors such as “slap”, “pat” due to the existence of intermediate touch behavioral patterns. Therefore a new classification approach for human touch behaviors using Self Organizing Map (SOM) is proposed, using its ability to classify the nearby data in to a group. Further in this research the axis of out put maps are defined to be meaningful by specifying the initial placement of the input data during the training process.