Implementing a Visual Assistant for Elderly Care using CCTV Camera

Please enable javascript in your browser.

Noori, Hassan | 2025

0 Viewed

Type of Document: M.Sc. Thesis
Language: Farsi
Document No: 58323 (05)
University: Sharif University of Technology
Department: Electrical Engineering
Advisor(s): Mohammadzadeh, Naejesolhoda
Abstract:
In this research, an intelligent and integrated system for elderly fall detection monitoring based on CCTV cameras has been designed and implemented. The system consists of three main components: a fall detection device, a central server, and a mobile application. The fall detection algorithm runs in real time on a Raspberry Pi 4B board and combines classical methods with deep learning techniques. Specifically, an SVM model is employed in the classical stage, while Convolutional Neural Networks (CNN) and MoveNet models are utilized in the deep learning stage to enhance accuracy. To ensure security and privacy, the system incorporates end-to-end encryption and secure HTTPS-based communication. The mobile application is developed using the Flutter framework and interacts directly with the Flask-based server. Performance evaluation on four benchmark datasets demonstrates that the proposed system achieved a sensitivity of 95% in detecting falls and a specificity of 95% in avoiding false detections on a final test set of 106 videos, while operating at speeds exceeding real-time processing requirements
Keywords:
Communication Security ; Computer Vision ; Video Analysis ; Machine Learning ; Telemedicine ; Eldery Care ; Elderly Health Monitoring ; Fall Detection

No TOC