Description

Skeleton-based Human Action Detection (HAD) has gathered increasing research interest in recent years. Unlike traditional Human Action Recognition (HAR), which classifies actions within pre-trimmed motion data, HAD presents a greater challenge by aiming to detect the start and end points of actions within untrimmed data. Furthermore, the majority of Skeleton-based HAD research has focused on healthy subjects, neglecting individuals with motion impairments. In fact, public motion datasets of impaired populations are scarce.

This project aims to develop an effective 3D skeleton-based action detection method for individuals with neuromuscular diseases, with potential applications in disease progression monitoring and robotic-assisted rehabilitation, among others.

The work is under the scope of the following projects:

• ArmTracker: A state-of-the-art wearable system to assess upper limb motor function in reallife conditions for patients with Duchenne muscular dystrophy and spinal muscular atrophy. (web)