A novel feature fusion technique for robust hand gesture recognition

Gesture recognition has emerged as a crucial field of research due to its wide range of applications, such as human-computer interaction, sign language interpretation, and virtual reality. In this research work, a robust framework for hand gesture recognition is developed. The publicly available dataset, i.e., the Multi-Modal Hand Gesture dataset, is used. To achieve effective classification of hand gestures, a series of preprocessing techniques are employed to enhance the image quality. The Haar wavelet transform (HWT) and Local Binary Pattern (LBP) techniques are utilized for feature fusion, extracting both global and local texture information. This fusion of features maximizes the discriminative power of the classification model. In addition, the original AlexNet model architecture is modified, now referred to as the AlexNet-11 Layer model, by incorporating additional layers that effectively learn the patterns and fused features. As a result, the proposed approach achieves a remarkable accuracy of 97.38%, outperforming the state-of-the-art methods, which achieved accuracies of 96.5%, 96% and 82%. This significant improvement showcases the effectiveness of the proposed approach in advancing the field of gesture recognition and image classification.