A Novel Approach to Hand Gestures Recognition Using Optimized Quantum GAN and ResNet-152 Features

The creation of hand gesture recognition approach, such as sign language applications, is crucial to bridging the communication gap with non-sign language users. This work proposes a novel approach to hand gesture detection with improved efficiency and accuracy through the use of sophisticated pre-processing, feature extraction, and classification algorithms. The suggested approach presents a robust image pre-processing technique called Window-aware guided image filtering, which smooths gesture images while maintaining edge details that are essential for feature extraction. After that, high-level discriminative features are extracted from gesture photos using ResNet-152, a deep convolutional neural network. In order to improve gesture classification even more, a Quantum Generative Adversarial Network (QGAN) is included. This network makes use of quantum computing to accelerate convergence and improve learning effectiveness. Furthermore, the Zebra Optimization (ZO) is utilized to optimize the QGAN model by fine-tuning the network parameters for enhanced recognition performance. This hybrid approach achieves superior accuracy, robustness, and speed, outperforming traditional hand gesture recognition systems. The presented approach gained the efficiency of around 99% for all recognition metrics.