The Metaverse, a virtual space combining VR, AR, and blockchain technologies, has the potential to revolutionize the healthcare and medical sectors. In telemedicine, the Metaverse allows doctors and patients to engage in real-time, high-quality interactions within a virtual space. In surgical simulations, not only medical students and surgeons but also nurses, clinical engineers, and other allied health professionals and their students—all current and future members of the medical team—can practice skills risk-free, thereby enhancing educational quality. Additionally, rehabilitation and fitness programs offered in the Metaverse encourage active patient participation in their treatment. However, its implementation faces technical, ethical, and economic challenges. Addressing these requires technological innovation, the establishment of legal frameworks, and financial support. Improving infrastructure, enhancing device accessibility, and ensuring data protection necessitate collaboration between governments, corporations, and academic institutions. By overcoming these obstacles, the Metaverse can significantly enhance healthcare services, providing high-quality care to a broader population.

The organic flexible electronics is an attractive research that wearable human sensing device on human fingernails. Generally, the strain gauge for human sensing is formed on plastic films. Therefore, when fix the film sensor on curved surface such as human nails, it is difficult good contact. In this study, we proposed a process which form patterned film of PEDOT:PSS (Poly(3,4-ethylenedioxythiophene):Poly(4-styrenesulfonate)) on 3D curved plastic curved substrate by micro contact printing method. Moreover, we fabricated strain gauge on flexible polyimide substrate and measured static electrical characteristics. It enables to form patterned film on such as human nails and fabricate strain gauge directly. The results show that direct printed PEDOT:PSS films on curved surface can be used to strain sensor, which leads to monitoring human body by sensing of human nail anytime.

We have developed blink measurement methods that can be applied to input interfaces. To use the eye-blinking information as an input trigger, it is necessary to automatically classify blink types into voluntary and involuntary. A method for blink type classification using a three-dimensional convolutional neural network (3D-CNN) has been proposed. This classification method takes a short image sequence of the periocular area and classifies the blink type. We previously reported on several performance-improving methods that can be applied to this 3D-CNN. Since our classification using 3D-CNN outputs classification results in units of video frames, multiple types of classification results could be mixed together during a period of a single blinking motion. To address this problem, we employ a correction method to calculate the mode value as a representative value for the consecutive blink period. This paper proposes a correction method to improve accuracy based on limiting the aggregation range of the mode to a reliable portion. The evaluation experiment resulted in 97.9% accuracy and 94.0% F-score in the classification results for each short image sequence for 10 subjects. Then, 97.5% accuracy and 97.3% F-score were obtained for the accuracy of blink type classification.

Skeletal muscle is a collection of motor units (MUs) composed of motoneurons and muscle fibres; muscle contraction is regulated by the number and type of MUs. Since sarcopenia and disuse atrophy are caused by atrophy of specific muscle fibre types, it is important to observe mobilized MUs for skeletal muscle diagnosis. In a previous study, we proposed a method to measure multi-channel surface electromyography (EMG) and quantitatively acquire conducting waves, which are potential waveforms propagating on the muscle surface. The mode of MU mobilization by prolonged muscle contraction is not well known. It is necessary to be able to estimate the mobilized MU over time to diagnose slow muscle fibres and to realize the evaluation of muscle activity during training. This study analyzed the conducting waves of multi-channel surface EMG during prolonged muscle contraction in 20 adult male subjects. The results of the changes in the conducting wave during long-duration muscle contractions at high and low loads showed different characteristics, suggesting the possibility of using the propagating wave to elucidate skeletal muscle.

In this paper, we propose a method for constructing a gaze direction estimation model using Vision Transformer (ViT) as a fundamental technology for developing input interfaces using eye movement. Eye movement measurement methods using machine learning enable highly accurate estimation but require many computational resources. Therefore, in this paper, we consider a calibration-free and lightweight gaze direction identification model with implementation in gaze input interfaces in mind. The proposed method constructs a gaze direction classification model by fine-tuning a large-scale pre-trained model of ViT using the constructed dataset. The training dataset was constructed by extracting the face region as a still image for each frame from a video image captured by a webcam and then focusing on the area near the eyeball. In addition, we experimented to evaluate the performance of the gaze direction estimation model constructed using the proposed method. As a result of the experiment, the accuracy rate and macro average F-value of the proposed method were approximately 19.0 points higher than the conventional method, and we confirmed the overall improvement of the classification performance under calibration-free.

Many healthcare-oriented menu recommendation methods are proposed in previous research. In some previous research, the purchase cost is minimized in addition to the minimization of the healthcare loss under the limit on the purchase cost when there are multiple optimal solutions. But in the previous research, the minimization of the healthcare loss considering meal history has not been studied. In this research, a recommendation method for restaurant menus that minimizes the healthcare loss under the limit on the purchase cost, considering meal history of family is proposed. The proposed method also calculates the minimum purchase cost when there are multiple optimal solutions on the healthcare loss. In the proposed method, the healthcare loss and the purchase cost are minimized by dynamic programming. The effectiveness of the proposed method is shown by some computational examples. In the computational examples, the healthcare loss of the proposed method is confirmed to be smaller than the loss of the empirical recommendation method. Adaptive menu selection examples under cost constraints, considering meal history of family are also confirmed. This research is basic research, and future extended research is required.

The location and area shape of neural activity sources were estimated using machine learning. In an inverse problem analysis to measure the magnetic field using a magnetic sensor and estimate the location of the current dipole (CD), eight magnetic sensors were used, and the percentage of location estimation errors of 1 mm or less was 43.2%, and the percentage of errors of 10 mm or more was 2.4%. In particular, the accuracy of location estimation was high on the z = 6 cm plane, and the percentage of errors of 1 mm or less increased. This shows high accuracy compared to the spatial resolution ability (5 to 7 mm), suggesting its effectiveness in estimating the area of neural activity sources.

Understanding the characteristics of micro gap discharge when applied surge voltage is important for insulation of miniaturized equipment. In this work, a microgap was created using a cone-shaped copper electrode, and an impulse voltage was applied. Light emission of micro gap discharge observed affected by rise time of voltage and investigated the effect on discharge behavior. It was found that the effect of shortening the rise time is greater when the gap length is large.

This research aimed to reveal the information process in the brain of the patients who suffered from mild cognitive impairment (MCI) using the power spectra and timelag analysis in electroencephalogram (EEG). The patients with MCI have memory or thinking problems, and are at a great risk of developing Alzheimer's disease or a related dementia. Ten patients with MCI and 13 healthy adults (non-MCI) were examined. The EEGs (122.88 seconds/subject) during a state of relaxation with the eyes closed, were analyzed using the Fourier analysis (power spectrum) and the timelag analysis. The results showed that the mean value of the power spectra in MCI was significantly higher than the mean value in non-MCI in the theta band. The mean value of the absolute timelag values (ATL) in MCI was significantly lower than the mean value in non-MCI in the frontal and left temporal areas. Our findings suggest that the features from EEG of MCI were abstracted in the theta band, and the propagation of information related to memory in brain would becoming fast in MCI.

Environmental monitoring in sea farms is crucial for detecting and predicting harmful signs in water quality for farmed organisms. While the use of unmanned surface vessels for this purpose has been considered, a single vessel is not necessarily able to meet the demands of frequent and high-density water monitoring requirements. This paper focuses on an implementing aspect of coverage control with unmanned trimaran ships for optimally deploying these ships to sea areas of interest. A coverage control law is implemented in a low-end microcontroller, and its practical viability and control performance is evaluated experimentally.

In this paper, we propose a classification method for “kawaii” images. Conventional image classification methods such as CNN (Convolutional Neural Network) and SVM (Support Vector Machine) can classify ordinary images such as faces with high accuracy, but they are insufficient for classifying images that are vaguely expressed without a “kawaii” explicit indicator.

Through experiments, we propose a suitable method for classifying “kawaii” images by extracting latent color, shape, and other features of “kawaii” images using feature filters, quantitatively representing them, and then comparing classification accuracy using various classifiers based on machine learning techniques.

In the experiments, the color, SIFT(Scale Invariant Feature Transform) and line features of the images are extracted using filters and compared using NN (Neural Network), Random Forest, AdaBoost and SVM. The results show the effectiveness of the proposed feature filters and the suitability of Random Forest as a classifier, and thus an effective method for classification of “kawaii” images.

In sports, various sounds are generated, such as the shouts of players, footsteps, and the sound of hitting the ball with a racket. In the sound of a bat hitting, differences in the intensity and pitch of the sound also appear with each batting. This paper describes an investigation and analysis of the relationship between the sound of a bat hitting and ball trajectory in baseball. In the experiments, datasets of the bat hitting sounds were constructed, and these hitting sounds were analyzed. Bat hitting sounds were recorded using a 32-bit float recording that enables recording without clipping. A total of 996 bat hitting sounds were recorded, and three types of datasets were constructed. Moreover, the ball trajectories were discriminated from a ball hitting sound using a support vector machine (SVM). In an experiment to discriminate three types of ball trajectories using SVM, an average correct rate was up to 48.4 %. The results suggest that it is possible to discriminate the ball trajectory from a bat hitting sound.

In this paper, we propose a classification method that deploys hard-to-explain rules and is robust against adversarial example (AE) attacks on the MNIST. The purpose of this paper is to solve the technical difficulties of the deep learning-based technology that include the unexplainability of the classification and the vulnerability against the AE attack. The method proposed in this paper is similar to the existing method (DkNN) in terms of using output vectors computed from an artificial neural network (ANN), thus can solve the unexplainability difficulty. The proposed method is different from the DkNN in terms of the architecture of used ANNs and the format of output vectors. Those output vectors are discrete and used as hard-to-explain rules that mitigate the vulnerability against the AE attack. In computational experiments, the MNIST is taken as the target problem, then FGSM and BIM are used as the AE attacks. Computational results display that the proposed method achieved accuracies over 95% for all attacks.

This paper proposes a data-driven nonlinear controller using Kolmogorov-Arnold Network (KAN). In contrast to a famous Multilayer Perceptron (MLP), KAN utilizes learnable b-spline curve as activation functions. Because of this, KAN can represent complex functions by small neurons, and has high interpretability. By employing KAN as a controller, the proposed controller has both high representational power and interpretability.

This paper proposes a Doppler radar-based method to estimate pedestrian height with a regression model using a convolutional neural network. Regarding the experiments with 16 adults (average height: 171.3 cm), the results showed that the mean absolute error for the height estimation was less than 2 cm. Furthermore, we found the leg motion information is more effective than the head and torso motions for height estimation.