Bonhak Koo

Measuring torsional movement is crucial for tracking human kinematics in applications such as virtual reality, biomechanics, rehabilitation, and soft robotic joints. However, continuous, non-invasive torsion measurement remains challenging due to multi-axial deformations, where existing methods rely heavily on empirical data fitting or computationally expensive deep learning models requiring vast training datasets. To address these limitations, this paper proposes a training-free, highly precise torsion estimation framework that converts soft resistive strain sensor data into forearm rotation angles using closed-form kinematic-geometric modeling. Two geometric representations—the Cylinder Model and Truncated Cone Model—are established to analytically map non-linear tensile strain to torsion angles. Additionally, an online calibration algorithm is integrated to mitigate cumulative measurement drift caused by viscoelastic hysteresis. Under static conditions, the proposed models achieved a high coefficient of determination (R² = 0.9888) against reference IMU data. In dynamic trials, the hysteresis compensation algorithm maintained robust tracking performance (R² = 0.8033), achieving an angular root-mean-square error (RMSE) of 4.12° and a cycle peak prominence relative error of 0.42% for the Truncated Cone Model. In this single-subject proof-of-concept feasibility study (n = 1), the proposed framework demonstrated the feasibility of continuous torsional motion tracking through a training-free geometric approach, providing a preliminary foundation for future evaluation in wearable systems and soft robotics.

The wrist is one of the most complex joints in our body, composed of eight bones. Therefore, measuring the angles of this intricate wrist movement can prove valuable in various fields such as sports analysis and rehabilitation. Textile stretch sensors can be easily produced by immersing an E-band in a SWCNT solution. The lightweight, cost-effective, and reproducible nature of textile stretch sensors makes them well suited for practical applications in clothing. In this paper, wrist angles were measured by attaching textile stretch sensors to an arm sleeve. Three sensors were utilized to measure all three axes of the wrist. Additionally, sensor precision was heightened through the utilization of the Multi-Layer Perceptron (MLP) technique, a subtype of deep learning. Rather than fixing the measurement values of each sensor to specific axes, we created an algorithm utilizing the coupling between sensors, allowing the measurement of wrist angles in three dimensions. Using this algorithm, the error angle of wrist angles measured with textile stretch sensors could be measured at less than 4.5°. This demonstrated higher accuracy compared to other soft sensors available for measuring wrist angles.

In recent years, human activity recognition (HAR) has gained significant interest from researchers in the sports and fitness industries. In this study, the authors have proposed a cascaded method including two classifying stages to classify fitness exercises, utilizing a decision tree as the first stage and a one-dimension convolutional neural network as the second stage. The data acquisition was carried out by five participants performing exercises while wearing an inertial measurement unit sensor attached to a wristband on their wrists. However, only data acquired along the z-axis of the IMU accelerator was used as input to train and test the proposed model, to simplify the model and optimize the training time while still achieving good performance. To examine the efficiency of the proposed method, the authors compared the performance of the cascaded model and the conventional 1D-CNN model. The obtained results showed an overall improvement in the accuracy of exercise classification by the proposed model, which was approximately 92%, compared to 82.4% for the 1D-CNN model. In addition, the authors suggested and evaluated two methods to optimize the clustering outcome of the first stage in the cascaded model. This research demonstrates that the proposed model, with advantages in terms of training time and computational cost, is able to classify fitness workouts with high performance. Therefore, with further development, it can be applied in various real-time HAR applications.