Multi-View Gait Recognition using CNN-GNN Fusion: A Deep Learning Framework

Gait recognition, a crucial biometric modality for human identification, is particularly important in security and surveillance applications. The challenge of multi-view gait recognition, which involves significant variations in viewing angles that can degrade recognition performance, is addressed by a novel model. This model proposes a Multi-View Gait Recognition Framework that leverages a fusion of Convolutional Neural Networks (CNNs) and Graph Neural Networks (GNNs). The model's significance lies in its ability to enhance cross-view gait recognition by integrating spatial and structural gait information. It uses CNNs to extract deep feature representations from gait silhouettes, followed by GNNs to establish inter-view dependencies and model structural correlations. The model is evaluated using publicly available gait datasets CASIA-B and OU-ISIR, demonstrating superior recognition accuracy and outperforming traditional CNN-based and handcrafted methods. The proposed CNN-GNN fusion model significantly improves cross-view consistency and robustness against intra-class variations, achieving accuracies of 98.2% and 96.7% on CASIA-B and OU-ISIR, respectively. This framework for CNN-GNN represents a promising direction for robust and efficient multi-view gait recognition.