Multi-Biometric Feature Extraction from Multiple Pose Estimation Algorithms for Cross-View Gait Recognition
Ausrukona Ray,Md. Zasim Uddin,Kamrul Hasan,Zinat Rahman Melody,Prodip Kumar Sarker,Md Atiqur Rahman Ahad
DOI: https://doi.org/10.3390/s24237669
IF: 3.9
2024-12-01
Sensors
Abstract:Gait recognition is a behavioral biometric technique that identifies individuals based on their unique walking patterns, enabling long-distance identification. Traditional gait recognition methods rely on appearance-based approaches that utilize background-subtracted silhouette sequences to extract gait features. While effective and easy to compute, these methods are susceptible to variations in clothing, carried objects, and illumination changes, compromising the extraction of discriminative features in real-world applications. In contrast, model-based approaches using skeletal key points offer robustness against these covariates. Advances in human pose estimation (HPE) algorithms using convolutional neural networks (CNNs) have facilitated the extraction of skeletal key points, addressing some challenges of model-based approaches. However, the performance of skeleton-based methods still lags behind that of appearance-based approaches. This paper aims to bridge this performance gap by introducing a multi-biometric framework that extracts features from multiple HPE algorithms for gait recognition, employing feature-level fusion (FLF) and decision-level fusion (DLF) by leveraging a single-source multi-sample technique. We utilized state-of-the-art HPE algorithms, OpenPose, AlphaPose, and HRNet, to generate diverse skeleton data samples from a single source video. Subsequently, we employed a residual graph convolutional network (ResGCN) to extract features from the generated skeleton data. In the FLF approach, the features extracted from ResGCN and applied to the skeleton data samples generated by multiple HPE algorithms are aggregated point-wise for gait recognition, while in the DLF approach, the decisions of ResGCN applied to each skeleton data sample are integrated using majority voting for the final recognition. Our proposed method demonstrated state-of-the-art skeleton-based cross-view gait recognition performance on a popular dataset, CASIA-B.
engineering, electrical & electronic,chemistry, analytical,instruments & instrumentation