Mitsunori Tada

Wearable exoskeletons have emerged as a solution to enhance locomotion in individuals with impairments and/or weakness. Assistive devices with pneumatic gel muscle actuators (PGMs) are promising for daily use due to their high power-to-weight ratio and compliant structure, enabling potentially easy integration into smart garments. The intrinsic properties of PGMs have been studied over the past decade; however, little is known about how to leverage their dynamics to effectively and optimally assist motion. In this study, we modeled hip joint assistance via two PGMs at each user’s leg and employed musculoskeletal simulations to predict optimal assistive strategies that reduce metabolic costs during walking at various speeds. Specifically, we implemented a bilevel optimization framework to identify optimal control parameters: stiffness, onset time, and duration, under two control modes: coupled and independent, at three actuator placements: medial, neutral, and lateral, relative to the user’s hip joint center. Our results showed that, across walking speeds, PGM actuators with coupled control mode reduced estimated metabolic cost by 5.3–16.0% and with independent control mode by 10.5–17.5%. We also identified that PGM assistance with medial placement with coupled control mode offered the best trade-off between control simplicity and potential metabolic savings at slow walking speeds, which might be particularly useful for enhancing mobility in older adults and in rehabilitation settings. Also, our simulation suggested that neutral placement tended to outperform other actuator placements across speeds in terms of metabolic savings. Future experimental studies may benefit from guiding exoskeleton control as per the predicted assistive strategies in this work.

Gait is a fundamental aspect of human mobility, and disruptions in normal gait can significantly reduce quality of life (QOL). Although recent advances in 3D gait analysis (3DGA) enable precise, quantitative assessments, these methods are typically confined to controlled laboratory environments and thus fail to accurately capture natural gait variability. Conversely, wearable IMU sensors offer cost-effective, portable solutions for capturing movements across diverse settings but face challenges such as invasiveness and sensor drift. In this study, we propose “Gait Inertial Poser (GIP),” a novel method estimating 3D full-body pose during straight walking on flat ground, using only two shoe-embedded IMU sensors. GIP initially estimates personalized body shapes from user attributes (height, weight, age, gender) and then employs a Transformer-based module to infer gait motion parameters from IMU data. To ensure temporal continuity and smoothness of the estimated motion, we further introduce a smoothing module based on a Variational Autoencoder (VAE), which further incorporates a specialized loss function that explicitly enforces kinematic constraints during foot-ground contact, thereby improving the overall estimation accuracy. Comprehensive experiments conducted on two public datasets quantitatively and qualitatively demonstrate that GIP achieves high accuracy in straight-line walking. This approach overcomes limitations of traditional laboratory-based methods, opening new opportunities for real-time monitoring and remote rehabilitation in everyday environments. The code will be available at https://github.com/RyosukeHori/GaitInertialPoser.

With the rapid technological advancements in wearable devices, motion and health management have significantly improved, enabling the measurement of various biometric data with compact equipment. Our research focuses on motion measurement but, in general, full-body motion estimation requires motion capture systems or multiple inertial sensors, making it necessary to directly measure movement itself. In this study, we propose estimating full-body posture using inverse kinematics based on trunk posture and limb-end information collected through wearable devices. To enhance estimation accuracy in this underdetermined problem, we employ Physics-Informed Neural Networks (PINNs), which efficiently learn using physical laws as a loss function, along with a high-precision inverse kinematics model of a digital human. Through this approach, we enable high-accuracy full-body posture estimation even with wearable devices in underdetermined scenarios.

In human behavior recognition using machine learning, model performance degrades when the training data and operational data follow different distributions which is a phenomenon known as domain shift. This study proposes a method for domain adaptation in the hidden semi-Markov model (HSMM) by modifying only the emission probability distributions. Assuming that the state transition probabilities remain unchanged, the method updates the emission probabilities based on the posterior distribution of the target domain. This approach enables domain adaptation with minimal computational cost without requiring model retraining. The effectiveness of the proposed method was evaluated on synthetic time-series data from different domains and actual care work data, achieving recognition performance comparable to that of models retrained for each domain. These findings suggest that the proposed method applies to various time-series data analysis tasks requiring domain adaptation.

In the quest for a blood pressure (BP) measurement method that is applicable in daily life, deep learning-based BP estimation models utilizing photoplethysmograms (PPGs) have recently been introduced. To allow routine deployment of BP estimation models that rely on wearable sensing PPGs, they should be able to sustain performance over an extended period. However, achieving stability is difficult due to variations in the distribution of PPG data caused by shifts in sensor positions and human conditions. This study assessed the performance of deep learning-based BP estimation models that were pretrained on a large-scale dataset and explored their adaptability and adjustment frequency during fine-tuning with wearable sensed small-scale data over a one-month period. Two distinct datasets were utilized: the publicly available Multiparameter Intelligent Monitoring in Intensive Care III Waveform Database Matched Subset version 1.0 and a private dataset recorded from 11 participants over four days (one day per week) using a customized wearable wristband-type PPG device. The results indicated the superior performance of the fine-tuned models compared with the non-fine-tuned models. Tuning the models every week resulted in greater performance than tuning only once in the first week. Balancing convenience and performance continues to be a practical challenge in this research field. Moreover, even the most accurate fine-tuned model does not consistently satisfy medical standards, presenting a hurdle for medical applications. This study revealed the effectiveness and adaptability of pretrained fine-tuned BP estimation models for long-term wearable sensing data; however, enhancing the estimation performance remains a challenge.

The potential of mental practice of motor tasks, activating motor brain areas through imagery to enhance brain plasticity and aid in patient recovery and rehabilitation, has garnered increasing interest. However, the most effective types of perceptual stimuli and presentation methods for promoting prehension movements remain unknown. In this research, we aimed to differentiate the effects of task-relevant aspects of visual stimuli, such as shape, from task-irrelevant aspects of visual stimuli, such as the orientation of graspable tools. Furthermore, we assessed the magnitude of these effects at different time points to clarify their temporal characteristics. Our experimental results revealed significant priming effects for both aspects of visual stimuli during the later phase but not the earlier phase, indicating engagement of diverse brain areas such as frontal lobe beyond the dorsal visual pathway in visual processing during prehension movements.