AUTOMATING THE ASSESSMENT OF WRIST MOTION IN TELEREHABILITATION WITH HAPTIC DEVICES

In my Master's project at NYU, I investigated the possibility of aiding stroke patients' rehabilitation through the use of a Novint Falcon, a desktop delta robot. This work, which also became the subject of a published conference paper, aimed to address the challenges of assessing motor performance in telerehabilitation due to the inability to directly measure the user's wrist angles while interacting with these robots. As an innovative solution, I tested the possibility of strapping a smartphone to patients' forearm as they manipulate the haptic robot. The data from the smartphone's inertial sensors and the robot were integrated and processed by a classification algorithm to infer the wrist angle.

In my Master's project at NYU, I investigated the possibility of aiding stroke patients' rehabilitation through the use of a Novint Falcon, a desktop delta robot. This work, which also became the subject of a published conference paper, aimed to address the challenges of assessing motor performance in telerehabilitation due to the inability to directly measure the user's wrist angles while interacting with these robots. As an innovative solution, I tested the possibility of strapping a smartphone to patients' forearm as they manipulate the haptic robot. The data from the smartphone's inertial sensors and the robot were integrated and processed by a classification algorithm to infer the wrist angle. The project was conducted in two significant phases. The first semester was dedicated to working with inertial measurement units (IMUs)—understanding their positioning, coding with Arduino, and carrying out an initial data collection. This groundwork provided a robust foundation for the next stage of the project. After the winter break, I pivoted towards Machine Learning. I developed algorithms in Python and conducted a second round of data collection. This phase culminated in the successful training of a machine learning model that predicts wrist angles from a single wearable sensor and the movements of the Novint Falcon. This approach represents a substantial stride towards the automatic assessment of wrist movements in fine motor telerehabilitation, potentially paving the way for real-time feedback without the need for a therapist's presence. The insights and conclusions derived from this project underscore the transformative potential of integrating robotics and machine learning in healthcare.