Year: 2023
Type: article
Abstract: Fall detection is vital for the elderly due to the prevalence of fall-related injuries. Implementing such systems can lead to timely interventions, better care, and reduced healthcare costs, improving the overall safety and well-being of older individuals. This paper proposes an innovative fall detection system that combines the Inertial Measurement Unit (IMU) approach with the computer vision approach. The IMU system uses the Random Forest algorithm (RF) for analyzing accelerometer data, distinguishing between normal activities and fall events. Concurrently, a computer vision (CV)-based approach uses Histogram of Oriented Gradients (HOG) feature extraction to train a Random Forest classifier, enabling fall detection based on visual cues. Further, the ensemble model combines predictions from both approaches and uses a voting classifier for final decision-making, leveraging the strengths of both approaches. Experiments were conducted on real-world datasets, demonstrating the high accuracy and sensitivity of detecting falls. The CV-based approach has shown 96% accuracy; the IMU-based approach has shown an accuracy of 98%; and the ensemble approach has shown 97% accuracy for fall detection. The implemented system significantly enhances fall detection, minimizes false alarms, and provides timely assistance for vulnerable individuals. It holds potential for improving fall detection systems, benefiting elderly individuals living independently.
Institution: Vishwakarma Institute of Technology
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: diamond
OpenAlex ID: W7216639827