Real-Time Hand Gesture Recognition System Using MediaPipe and Deep Learning for Smart Control

This paper presents a real-time hand gesture recognition system using MediaPipe and deep learning techniques for smart control applications. The system extracts 21 hand landmarks from RGB video streams and applies feature engineering to compute normalized coordinates, distances, and joint angles. These features are classified using a Multi-Layer Perceptron (MLP) model to recognize various gestures efficiently. The proposed system is lightweight and operates in real-time on CPU without requiring specialized hardware. Experimental results show an accuracy of approximately 94–95% under varying lighting conditions and backgrounds. The system demonstrates robustness, scalability, and suitability for applications such as human-computer interaction, smart home control, virtual presentations, and assistive technologies.