Smart Camera-Based Real-Time 2D Arm Gesture Detection: A Fuzzy Logic and CNN Approach

In today's technological era, natural remote control of devices is essential for effortless system management from a distance. However, many current solutions are complex and require advanced skills, limiting user accessibility. This study presents Smart Camera Two (SCT), an advanced system enabling precise real-time recognition of upward and downward arm gestures on both arms, with a minimum angular resolution of 15°. Our research assesses the performance of the SCT in its two variants (SCT-O and SCT-T) compared to the earlier SCO, considering factors and materials used [1]. The system uses a camera that faces the user to capture simple arm movements, enabling intuitive interaction without technical expertise. This natural gesture control streamlines tasks and significantly reduces completion times. Real-time data acquisition facilitates prompt decision-making and adaptive system monitoring, boosting productivity. SCT achieves an overall accuracy of 99.937%, with low latency and robustness against lighting changes and occlusions. It combines two versions: SCT-O employing Fuzzy Logic (FLO) and SCT-T based on convolutional neural networks (CNNs), each suited to different project needs. This article presents a detailed comparative analysis of both versions, supported by extensive experiments that validate their performance. A video presentation of this research is publicly availble https://sites.google.com/view/smart-camera-two-sct/accueil.