In this research, we examined the differences between different gazing positions during training. We used multiple electrodes and focused on the amplitude and conduction velocity of the measured conducting waves, which enabled a detailed examination of the muscle contraction mechanism. In this experiment, the test muscle was the biceps brachii muscle, which was trained under three separate conditions to investigate differences in conducting wave characteristics depending on gazing position. As a result, no differences were found in the relative frequency distribution or in the percentage of the frequency of the occurrence of the antiphase conducting wave pairs due to the difference in the gazing position. However, it was shown that the ratio of the frequency of the occurrence of the antiphase conducting wave pairs could be used as an indicator of the training effect.

Blood viscosity, which is an important index during the extracorporeal treatment, is one of the determinants of red blood cell aggregation in the blood. In this paper, we propose an optical detection method for red blood cell aggregations by using object detection algorithms, assuming that they will be incorporated into the flow path of an extracorporeal circulation device. By performing detection from a single image, it is a method that can detect aggregations without being affected by rotation, scaling, or the attachment and detachment of red blood cells. In the simulated red blood cell sample, the detectability of aggregations was shown not only at the focus site but also at the defocus site.

Human factors account for the majority of traffic accidents. Identifying and preventing “attempted accidents” as one example, is anticipated to reduce the overall incidence of traffic accidents. Effective detection of attempted accidents requires an evaluation of driver's physiological and psychological states, with stress coping styles serving as a key method of assessment. When exposed to stress, individuals exhibit two primary stress-coping patterns, which could be distinguished based on hemodynamic indices measurable with a continuous blood pressure monitor. This approach has been widely applied in studies of human physiological and psychological responses. Additionally, it has been suggested that these stress-coping patterns may shift in response to driver interactions during vehicle operation, indicating their potential utility in evaluating driver's physiological and psychological states. However, continuous blood pressure monitor presents safety concerns, as the equipment constrains driver movement. To address this, the present study explores the feasibility of inferring stress-coping styles through unobtrusive measurements of driving behaviors alone. We analyzed driver's behavior and hemodynamic indices during the driving segment immediately preceding a segment with a potential attempted accident, aiming to elucidate differences in driving responses with and without the anticipation of an accident across various psychological states.

This study focuses on a multi-hop control network (MHCN), where a single controller aims to control multiple plants through a common multi-hop network. In the focused MHCN, the communication delays occur, and the length of the communication delay depends on the number of relays in the selected route. Thus, it is necessary to consider that the timing for optimization at the controller may vary for individual plants. A synchronous optimization method has been proposed to simultaneously determine routes and control inputs to multiple plants, using the time of the maximum communication delay as the period for optimization. However, using this method, even if a route with a short communication delay is selected, the benefit of selecting a route with a small communication delay is diminished because the next control is performed only after waiting for the maximum delay. Therefore, this paper proposes an asynchronous optimization method that jointly optimizes the communication route and control input upon receiving the control state, enabling immediate optimization of the corresponding plants. This method leading to efficient utilization of the multi-hop network. Furthermore, we extend the proposed asynchronous optimization method that takes into account disturbances whose impact depends on the communication delay.

Utility pole deterioration has been diagnosed by performing a percussion inspection after visual inspection of longitudinal cracks on the pole surface. In recent years, with the increase in the number of typhoons that come ashore with increasing force, there have been many closed transverse cracks on the pole surfaces caused by strong winds. Although abnormality detection using Artificial Intelligence has been studied for deterioration diagnosis in recent years, it is difficult to collect a large amount of data on healthy poles immediately after erection, and even more difficult to collect data on poles that are similarly deteriorated. Therefore, an experiment was conducted using an undamaged pole immediately after erection and a pole with closed cracks on the surface after a bending test. As a result, we improved the dependence of the resonance frequency of the ultrasonic probe using white noise, and by using short-time Fourier transform images of the propagation frequency, we were able to detect the deterioration of the poles due to the occurrence of closed cracks.

Social isolation and loneliness among older adults have become important social and public health issues. Such a situation will make it difficult for older adults to stay motivated for community activities such as walking. This study aims to incentivize older adults to engage in regular social activities and promote self-motivated behavioral changes. Therefore, this study introduces two ranking systems based on equality and equity in multi-agent simulation, which aim to promote fair interactions between older adults and young people and encourage older adults to engage in community activities and social contact. This study demonstrated the effectiveness of the psychological incentive design through both simulation experiments and field experiments.

This study investigates how dynamic psychological mechanisms—particularly self-efficacy—can enhance user engagement in community-based digital systems. We combine a multi-agent simulation and a 121-day field experiment to compare system designs with and without psychological adaptation. In the simulation, agents with adaptive self-efficacy formed larger social networks and exhibited more gift-giving behavior. In the real-world setting, three staged interventions—a gift reimbursement scheme, social rankings, and a step prediction contest—were introduced to stimulate motivation. Results show that while the effect of extrinsic incentives declined over time, user participation remained stable, suggesting partial internalization of motivation. The study provides empirical evidence for integrating psychological constructs into digital community design and offers a framework for measuring motivational impact beyond short-term reward effects.

In optimization of so-called black-box systems with unknown input-output relationships, we proposed the concept of "modeling-driven optimization," which simultaneously constructs a surrogate model of the black box and identifies the optimal input-output of the model for a given objective function, and presented the problem formulation and solution method in Reference (1). That study assumed that a sufficient number of input samples for modeling were taken from a region containing the unknown optimal solution. In contrast, this paper considers cases where, similar to fine-tuning in machine learning, a small number of input samples are taken from a relatively narrow region, and the optimal solution is not necessarily contained within that region. For such cases, we formulate the problem of determining the input sample region of the modeling-driven optimization problem so that it contains the unknown optimal solution as a bilevel optimization problem. As a solution method for this bilevel optimization problem, we propose a computational method that treats the input sample region as a hyperparameter of the lower-level modeling-driven optimization problem. This approach searches for the optimal sample region by considering the solution of the modeling-driven optimization problem. We confirm the effectiveness of the proposed method through a simple example.

PYNet estimates the ground surface of an object as a classification task. This makes pose estimation easier and allows an RGB-D camera to estimate poses with high accuracy. However, if we use only one viewpoint, the accuracy may decrease depending on the relationship between an object's pose and the viewpoint. In this study, we propose a method that combines multi-view images effectively by using classification scores of ground surface. The classification scores from each viewpoint are used as confidence values. We use these values to give different weights to the features from each viewpoint for pose estimation. In our experiment, the rate of pose estimation within 5 degrees was 46.4% for CosyPose, a well-known multi-view method. On the other hand, our proposed method achieved 61.5%, improving the rate by 15.1 points.

This paper proposes a play recording user interface (UI) for a table tennis match video with two key features: rally skip playback and swing loop playback. These functions aim to improve the efficiency and accuracy of manual play recording, which currently requires repeated video reviews. The rally skip uses image processing to estimate rally start times, allowing users to jump directly to relevant scenes. In the swing loop feature, swing moments are identified by analyzing the ball's movement over the table, enabling repeated playback of specific swings for precise input.

In comparative experiments with a baseline method, the rally skip function was the fastest, with the average recording time being approximately 1.3 times the length of the video. Furthermore, in experiments concerning recording accuracy, the swing loop function showed the best accuracy for strokes after the second shot, while the rally skip function showed the highest accuracy for other items. User feedback noted issues such as missing swing timestamps due to detection errors and difficulty tracking swings in long rallies. Future improvements include enhancing swing detection and adding on-screen indicators for swing numbers and player positions. These results show that the proposed UI effectively enhances both the efficiency and precision of play recording, offering a valuable tool for table tennis performance analysis.

Deep learning is a de-facto in computer vision. MLP-Mixer is a deep learning-based image classifier model that does not have any convolutional or self-attention mechanism, and is attracting attention for its competitiveness of accuracy and throughput against conventional convolutional neural networks (CNNs) and state-of-the-art models such as Vision Transformer (ViT). This study proposes a binarized MLP-Mixer aimed at implementation to IoT edge devices, and evaluate its effectiveness through experiments. The results demonstrate that MLP-Mixer can be trained with binarized weights, and when activations were 4 and 8bits, accuracy of 0.9 or higher was achieved on the MNIST handwritten digits dataset.

In this letter, an equivalent circuit representation of a 1/4 wavelength uniform transmission line consists of five lumped elements is presented, and a tunable power divider using the equivalent circuit is proposed. This divider can be varied the operating frequencies. The good frequency characteristics could be gotten by using circuit simulator.

In this study, we aimed to verify the reproducibility of a previous report suggesting that extremely low-frequency electric fields (ELF-EFs) promote calcification in differentiated MC3T3-E1 cells. A follow-up experiment was conducted using a system engineered to produce the most spatially uniform 60 Hz EF achievable under experimental conditions, which constituted a feature of this study. The exposure conditions were 3 hours per day for 9 consecutive days, and calcium deposition was quantified using alizarin red staining. As a result, a significant enhancement of calcification was observed in the ELF-EFs exposure group, consistent with previous findings. These results reinforce the hypothesis that exposure to ELF-EFs enhances mineralization in osteoblast-like cells.

The CSP (Cross-power Spectrum Phase) method is well-known as a source localization method. This method often estimates a noise direction under a stationary colored noise environment. On the other hand, the sub-band CSP method, which is a speech direction enhancement method, gives the noise direction as the estimation result when the noise power is dominant at low frequencies. Therefore, it is effective to develop an estimation method based on a difference of properties on the time axis between speech signals and noise signals. In this paper, a novel method based on correlation value weightings is proposed. The effectiveness of the method is shown by experimental results in a room environment.

Combinatorial Optimization Method Based on Hierarchical Structure in Solution Space (HS-COM) has been shown to outperform Tabu Search. In this letter, we identify cycling of search points as a factor degrading performance of HS-COM and propose an improved approach that dynamically adjusts parameters to address this issue. Numerical experiments on benchmark problems show that the improved method outperforms the original.