KINECAL: A Dataset for Falls-Risk Assessment and Balance Impairment Analysis

Area of Science:

• Biomedical Engineering
• Human-Computer Interaction
• Clinical Biomechanics

Background:

• Human action recognition has advanced due to motion capture datasets, particularly those using Kinect.
• Clinical applications of Kinect are limited by a lack of specialized datasets with relevant clinical movements and metadata.

Purpose of the Study:

• To introduce KINECAL, a novel dataset designed to bridge the gap between motion capture technology and clinical practice.
• To provide researchers with clinically relevant data for developing and validating human action recognition algorithms in healthcare settings.

Main Methods:

• The KINECAL dataset comprises recordings of 90 individuals performing 11 distinct movements.
• Movements included are commonly used in clinical balance assessments.
• The dataset is enriched with metadata, including clinical labels, falls history, and postural sway metrics.

Main Results:

• KINECAL offers a comprehensive collection of motion capture data for balance assessment.
• The integrated metadata provides valuable context for clinical analysis and algorithm development.
• The dataset facilitates research into the clinical utility of motion analysis.

Conclusions:

• KINECAL addresses the need for clinically relevant motion capture datasets.
• This resource is expected to accelerate research in clinical motion analysis and the application of Kinect technology in healthcare.
• The dataset supports the development of advanced human action recognition systems for clinical use.