Deep multi-convolutional stacked capsule network fostered human gait recognition from enhanced gait energy image

Gait recognition is a well-known biometric identification technology and is widely employed in different fields. Due to the advantages of deep learning, such as self-learning capability, high accuracy and excellent generalization ability, various deep network algorithms have been applied in biometric recognition. Numerous studies have been conducted in this area; however, they may not always yield the expected outcomes owing to the issue of data imbalance in clinical and healthcare industries. To overcome this problem, deep multi-convolutional stacked capsule network fostered human gait recognition from enhanced gait energy image (HGR-DMCSCN) is proposed in this manuscript. Initially, the input images are taken from CASIA B and OU-ISIR datasets. Then the input images are given to preprocessing segment to enhance the superiority of the images based upon contrast-limited adaptive histogram equalization filtering (CLAHEF). Then preprocessed image is given to classification process using deep multi-convolutional stacked capsule network (DMCSCN) that is utilized for human gait detection under various conditions, like normal walking, carrying a bag and wearing a cloth. The proposed HGR-DMCSCN approach is executed in python and its performance is examined under performance metrics, such as F-Score, accuracy, RoC and computational time. Finally, the proposed approach attains 28.70%, 11.87% and 14.79% higher accuracy for CASIA B compared with existing methods.