Efficient Gait Recognition Using a CNN-LSTM Framework Optimized with Hippopotamus Algorithm

Abstract

Gait recognition has emerged as a vital biometric technique for unobtrusive human identification in surveillance, healthcare, and behavioral analytics. However, achieving high accuracy under real-world variations such as walking speed, clothing, and viewpoint changes remains a significant challenge. This study proposes an advanced gait recognition framework that combines a Convolutional Neural Network (CNN) for spatial feature extraction with a Long Short-Term Memory (LSTM) network for modeling temporal dynamics. To address the limitations of manual hyperparameter tuning, the Hippopotamus Optimization Algorithm (HOA); a bio-inspired metaheuristic is integrated to optimize key parameters such as learning rate, filter size, LSTM units, and dropout rate. The model is evaluated on the TUM GAID dataset, enco­­­­­mpassing diverse gait variations. Experimental results demonstrate that the HOA-optimized CNN-LSTM architecture significantly outperforms baseline and state-of-the-art methods in terms of recognition accuracy, Genuine Acceptance Rate (GAR), and Equal Error Rate (EER). The proposed framework exhibits superior robustness and convergence speed, affirming the efficacy of metaheuristic-driven optimization in deep learning-based gait recognition systems.