Automatic Detection of Protective Movement Behavior with MoCap and sEMG Data for Chronic Pain Rehabilitation

Physical rehabilitation is an important part of chronic pain (CP) management. Physiotherapists provide guidance and intervention on the exercises based on the CP patient's affective states. One important clue the physiotherapist uses to understand their patient's affective state is the presence and type of protective movement behavior. As rehabilitation is transferring from clinical settings to home-based environments, technology should provide similar service by automatically detecting the protective behavior exhibited by patients and use it as a cue to inform and adapt the support. Our research focuses on the detection of protective behavior from a deep learning (DL) perspective and using MoCap and EMG data. Based on the knowledge learned from a wider-relevant literature and the specific characteristic of protective behavior, we aim to automatically detect protective behavior with deep learning approaches and further learn its configuration pattern with explainable models. Our initial studies have demonstrated interesting accuracy improvements and also provided important knowledges about the temporal and configurational characteristics of protective behavior.