GESTURE AND FACE RECOGNITION SUBSYSTEMS, AND SYSTEM INTEGRATION IN HOME SECURITY LEVEL IDENTIFICATION SYSTEM BASED ON BEHAVIOR DETECTION

• Main Author: Faris Muhammad, Dafa
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/41805
• Institution: Institut Teknologi Bandung

The ever increasing of computational power means there’re more daily live aspects which can be replaced by computer. One of the most important aspect with fast development because of this is in security systems, with vastly more information which can be obtained without even needing any continuous human interaction. Using pose estimation, a system capable for recognizing body gestures is designed. A deeper level of recognition then achieved with the help from various advanced data processing, thus even going as far as giving an info with high abstraction level such as how secure a room is. The addition of face recognition is used for controlling the authority of someone in doing something, while object detection increases recognition variables. The testing is divided into two, the system itself and the minimum hardware requirements to run the system. In test scenario, the system has reliability to cover at minimum 88% of its initial coverage when one of four of its cameras is dead, face recognition accuracy is 90.44%, and final room security level decision accuracy is 93.75%. Meanwhile the hardware used must be at least capable of running the system at 3 FPS, which is tested to be attainable by using Nvidia GPU with minimum Compute Capability of 3.5 and minimum VRAM of 4 GB.