A Hand Gesture Recognition Model Using Deep Learning Algorithms
Janhavi Rajendra Kale, Rajani Mahesh Yemul, Shubham Suresh Vyavahare, L. M. R. J. Lobo, Vijay Anant Athavale · Advances in computational intelligence and robotics book series · 2025
Hand Gesture Recognition (HGR) is transforming human-computer interaction by offering a natural communication method. This study presents a robust HGR model using Deep Learning and CNNs to capture spatial hierarchies and complex visual patterns. The system processes RGB image/video inputs with varied lighting, backgrounds, and orientations. Preprocessing includes resizing, normalization, background subtraction, histogram equalization, and data augmentation. The CNN architecture uses fully connected layers with ReLU and softmax activation. It classifies gestures with real-time outputs and confidence scores. Evaluated on datasets like ASL, the model showed high accuracy, recall, precision and F1-score, outperforming traditional methods. It performed reliably in real-time leveraging transfer learning with networks like VGGNet, ResNet and MobileNet. Its lightweight design supports edge deployment. This work demonstrates the effectiveness of CNN-based HGR and suggests future research in attention mechanisms, multimodal data, and dynamic gesture recognition for adaptive systems.