Dynamic Hand Gesture Recognition with 2DCNN-LSTM and Improved Keyframe Extraction

This paper presents a novel approach to dynamic hand gesture recognition that surpasses existing methods in accuracy and efficiency. By combining deep learning and clustering techniques, we propose a new framework for keyframe extraction that effectively captures representative video frames. Furthermore, a novel data augmentation method is introduced to enhance the robustness of our proposed 2DCNN-LSTM model. Our model achieves a state-of-the-art accuracy of 98.69% on the Northwestern University Hand Gestures Dataset, demonstrating the effectiveness of our approach in recognizing complex hand gestures. This research enhances the advancing human-computer interaction field by providing a robust and accurate method for dynamic hand gesture recognition, with potential applications in virtual reality, augmented reality, and human-robot interaction.