BiGaitNet: A Hybrid Approach for View-Invariant Gait Recognition

Human gait is a fascinating and unique biometric feature that can tell us a lot about individuals, making it an invaluable tool for applications in security, surveillance, and even human-computer interaction. Being able to recognize people by their walking patterns opens up new possibilities for enhancing safety and improving our interactions with technology in everyday life. In this paper, we introduce BiGaitNet, an innovative hybrid model that combines the strengths of Convolutional Neural Networks (CNN) and Bidirectional Long Short-Term Memory (BiLSTM) networks for robust gait recognition. Our approach utilizes silhouette data captured from various perspectives—Lateral, Oblique, and Frontal—ensuring that the model performs well regardless of how the person is viewed. Through extensive testing, BiGaitNet achieved an impressive test accuracy of 96.35%, demonstrating its effectiveness in identifying individuals based on their unique walking styles. These promising results position BiGaitNet as a powerful tool for biometric identification, paving the way for advancements in how we understand and utilize human gait in technology.