AI_home security system

Conventional camera-based monitoring systems commonly require a person to watch video feeds or review recordings after an event. This paper proposes a laptop-based AI Smart Security and Activity Monitoring System that analyzes live video from a USB camera, combines person detection, face matching, and selected activity analysis, and sends qualifying events to a security decision engine. The engine can record event metadata, display notifications, and activate a USBconnected siren when the device supports software control. The proposed architecture uses Python, OpenCV, a lightweight YOLO-family detector, a face-detection and embedding-based recognition component, tracking, and temporal activity rules. PySide6 and SQLite are proposed for the local dashboard and event history. The design prioritizes local processing, modularity, low additional hardware cost, and clear separation between AI predictions and alarm decisions. Because the system has not yet been experimentally evaluated, no accuracy, latency, or reliability results are claimed. A validation plan is provided using precision, recall, false-alarm rate, processing latency, and sustained-run stability. The paper also discusses limitations arising from lighting, occlusion, face-matching uncertainty, computational load, privacy, and the USB siren's control interface.