Ms. Jumana Jabin, Ms. Lemya Sainudeen
The increasing elderly population worldwide has raised significant concerns regarding their safety, especially for those living independently. Falls, medical emergencies, and prolonged inactivity often go undetected, leading to severe health consequences. Existing monitoring solutions such as wearable devices and camera-based systems suffer from limitations including user non-compliance, high cost, and privacy invasion. This paper presents SafeSound, an intelligent and non-intrusive elderly safety monitoring system that utilizes on-device artificial intelligence to analyse ambient audio signals. The proposed system integrates three major functionalities: fall detection through sound pattern recognition, voice-based emergency command detection, and inactivity monitoring based on environmental sound levels. The system employs lightweight machine learning models optimized for mobile devices using TensorFlow Lite, ensuring real-time processing without transmitting raw audio data, thereby preserving user privacy. Experimental design and implementation demonstrate that the system can effectively detect emergency events and provide immediate alerts to caregivers through a cloud-based notification system. The proposed approach offers a cost-effective, privacy-preserving, and user-friendly solution that enhances elderly independence while ensuring rapid response in critical situations.