Valenti, Emanuele, Piscetta, Silvia, Ye, Yutao, Qureshi, Mehak Fatima, Kango, Rizwan Ahmed, Umair, Zuneera, and Qureshi, Umair Mujtaba (2025) Context-Aware Automation: AI Based Mood Detection for Personalized Smart Homes. In: Proceedings of the 8th International Conference on Communication Engineering and Technology. pp. 109-114. From: ICCET 2025: 8th International Conference on Communication Engineering and Technology, 16-18 May 2025, Guangzhou, China.

PDF (Published Version) - Published Version Restricted to Repository staff only

Abstract

Smart homes have become a promising way to enhance the comfort and convenience of living environments. The use of artificial intelligence (AI) technologies in smart homes has transformed smart homes by enabling automated systems that learn user preferences, optimize energy usage, and enhance security for a more convenient and efficient living environment. In addition, these technologies work together to create context-aware solutions by leveraging real-time data from connected devices to understand user behavior and environmental conditions. This integration enables systems to adapt dynamically, offering personalized experiences, optimizing resource usage, and enhancing decision-making. One example of such systems is the use of human moods or expressions to automatically adjust smart home systems based on state-of-the-art AI based Mood detectors. In this research, we have developed an AI-based mood detection system using Google's MediaPipe. The application can automatically analyse the current of mood of the user (happy, sad, angry or neutral), and adjust the surroundings based on his current mood such as playing the music and sharing contextual texts. The results show that our application model performs very well with an average accuracy up to 69-70% while testing on the images of publicly available dataset, and also predicts the user's mood in real-time. Furthermore, to secure our application we have implemented the strong authentication mechanism for facial recognition and password protected system.

Actions (Repository Staff Only)

Item Control Page