A Human Motion Data Capture Study The University of Liverpool Rehabilitation Exercise Dataset.

Area of Science:

• Biomedical Engineering
• Computer Science
• Rehabilitation Science

Background:

• Telerehabilitation utilizes motion tracking for remote exercise interventions.
• Human Action Recognition (HAR) research relies on high-quality human motion data.
• A significant gap exists in publicly available human motion datasets for telerehabilitation.

Purpose of the Study:

• To introduce the University of Liverpool Rehabilitation Exercise Dataset (UL-RED).
• To provide a comprehensive dataset for advancing HAR in telerehabilitation.
• To address the need for diverse human motion data, including varying motion speeds.

Main Methods:

• Collected data from 10 subjects performing 22 non-specialised exercises.
• Utilized three data modalities: marker-based motion tracking, marker-less motion tracking, and depth data.
• Recorded exercises at normal, fast, and slow paces to capture motion variability.

Main Results:

• The UL-RED dataset comprises 1,320 recordings across three modalities.
• Includes over three hours of marker-based and marker-less motion tracking data.
• First dataset to incorporate varying motion speeds for rehabilitation exercises.

Conclusions:

• The UL-RED dataset supports the development and evaluation of HAR systems for telerehabilitation.
• Facilitates advancements in remote exercise monitoring and analysis.
• Offers a valuable resource for both telerehabilitation and broader HAR research fields.