Gait Identification using Ensemble Deep Learning Technique

Gait is a distinctive non-invasive biometric feature that may be used to reliably identify people, even if they are unwilling to cooperate. Identifying people based on their strides is an emerging area in computer vision. Gait recognition still confronts significant obstacles in practical applications due to the intricate outside factors that were taken into account while choosing the gait data sample and the recognized individual’s attire changes. In this work, a deep neural network ensemble model is proposed to address the the gait identification issue. In ensemble learning several deep neural networks may be trained to carry out the same task very well. To acquire deep features, we first employ two convolutional neural networks (CNNs) as feature extractors for gait energy images (GEI) and motion silhouette images (MSI). Secondly, the features recovered from the two networks are fused to boost gait identification reliability. Finally, given the fused features, we utilize the softmax classifier for final prediction. CASIA B dataset is used for the assessment of the proposed method. The study shows us that the ensemble network works well when compared with single networks and outperforms a single deep neural network, in terms of generalization.