Towards Personalized Interaction and Corrective Feedback of a Socially Assistive Robot for Post-Stroke Rehabilitation Therapy

摘要: A robotic exercise coaching system requires the capability of automatically assessing a patient’s to in-teract with patient and generate corrective feedback. However, even if patients have various physical conditions, most prior work on systems has utilized generic, pre-defined feedback.This paper presents an interactive approach that combines machine learning rule-based models assess rehabilitation tunes data personalized To feedback when erroneous motion occurs, our applies ensemble voting method leverages predictions from multiple frames for frame-level assessment. According evaluation dataset three stroke exercises 15 post-stroke subjects, supports more accurate assessment (p < 0.01), but also can be tuned held-out user’s unaffected motions significantly improve performance 0.7447 0.8235 average F1-scores over all 0.01). This discusses value interaction system.