Authors:
Chettiyar Vani Vivekanand, M. Pavithra, S. Kalaivani, Ciyamala Kushbu, E. Anna Devi, D. Samson

Addresses:
Department of Computer Science and Engineering, Rajalakshmi Engineering College, Chennai, Tamil Nadu, India. Department of Electronics and Communication Engineering, Rajalakshmi Institute of Technology, Chennai, Tamil Nadu, India. Department of Information Science and Engineering, CMR Institute of Technology, Bengaluru, Karnataka, India. Department of Electronics and Communication Engineering, Sathyabama Institute of Science and Technology, Chennai, Tamil Nadu, India. School of Computer Science and Information Technology, University of Edenberg, Lusaka, Zambia.

Abstract:

Since falls are a leading cause of harm and diminished mobility in the elderly population, there is a need for discreet and dependable indoor monitoring systems. Current vision-based methods primarily focus on post-fall detection and often exhibit poor adaptability to individual movement patterns and high false alarm rates. This paper suggests a vision-based, real-time, non-wearable fall risk prediction system that incorporates a human-in-the-loop confirmation mechanism to improve practical usability and proactively estimates fall risk. The system continuously analyses human posture using a single monocular RGB camera and MediaPipe-based skeletal pose estimation. To account for inter-subject variability, an online, personalized baseline posture model is learned without prior calibration. A continuous scoring method based on torso angle deviation and temporal instability features is used to quantify fall risk. Before alert generation, a temporal instability observation window is used to minimize false alarms. Additionally, users can turn off pointless alerts with an audio prompt and gesture-based confirmation that uses right-hand raise detection. The suggested system is suitable for proactive elderly safety monitoring, as evidenced by experimental results from real-time indoor video streams demonstrating reliable fall risk tracking, fewer false alarms, and real-time feasibility.

Keywords: Fall Risk Prediction; Elderly Safety Monitoring; Vision-Based System; Human Pose Estimation; Real-Time Monitoring; Temporal Instability Analysis; Gesture-Based Confirmation.

Received on: 13/05/2025, Revised on: 24/07/2025, Accepted on: 03/09/2025, Published on: 09/03/2026

DOI: 10.69888/FTSBE.2026.000669

FMDB Transactions on Sustainable Biomedical Engineering, 2026 Vol. 1 No. 1, Pages: 1-17

• Views : 29
• Downloads : 8

Download PDF