Hall E-Smart palm sensor for autonomous home-based hand rehabilitation

Abstract

Hand rehabilitation is an important process for restoring functionality and quality of life, targeting injuries such as "avocado hand", with tendon or nerve damage. In such applications, estimation of finger forces plays a vital role in understanding tactile interactions and hand function. This thesis presents the design and development of a tactile sensor plate based on Hall-effect sensors, specifically the MLX90393, for detecting and analyzing finger forces and gestures. The primary goal is to design a sensor plate able to perform precise measurement of magnitude and directions of the forces exerted by the fingers, providing real-time feedback to users on 3D force distribution, contact areas, and tactile gestures to enhance interaction analysis and control. The tactile sensor plate design includes a matrix of 48 Hall-effect sensors, and one Inertial Measurement Unit, embedded within a compact, silicone-coated plate designed for comfort, durability, and ease of cleaning. By collecting sensor data at a sampling rate of 100 Hz, the plate is capable of estimating both normal and shear forces, important for understanding and guiding finger movements. A machinelearning Random Forest model was developed to process sensor data and predict applied force parameters. The results demonstrate the sensor plate’s ability to measure force magnitudes and directions with high accuracy, achieving an R2 value of approximately 96% − 99% for force prediction in specific test cases. The research addresses challenges such as hysteresis caused by the silicone layer, calibration for diverse hand sizes, and environmental conditions like magnetic field variations. The system architecture, built on a Raspberry Pi 5 as a main processing unit, opens possibilities for future enhancements, including the integration of additional sensors and improved calibration methods. This study contributes to the field of tactile sensing and force measurement by offering a precise detection of finger movements. The proposed system can estimate the applied force making a foundation for supporting guided exercises, allowing a better recovery process, and reducing dependence on physiotherapists. It also aligns with sustainable design principles. Future work will focus on enhancing sensor accuracy, refining the graphical user interface, and exploring potential applications in various tactile sensing and human-machine interaction scenarios.