Human gait analysis for osteoarthritis prediction: a framework of deep learning and kernel extreme learning machine

Abstract Human gait analysis is a novel topic in the field of computer vision with many famous applications like prediction osteoarthritis and patient surveillance. In this application, abnormal behavior problems walking style detected suspected patients. The means assessments terms knee joints any other symptoms that directly affected patients’ style. carries substantial importance medical domain, but variability clothes, viewing angle, carrying conditions, may severely affect performance system. Several deep learning techniques, specifically focusing on efficient feature selection, have been recently proposed for purpose, unfortunately, their accuracy rather constrained. To address disparity, we propose an aggregation robust features Kernel Extreme Learning Machine. framework consists series steps. First, two pre-trained Convolutional Neural Network models are retrained public datasets using transfer learning, extracted from fully connected layers. Second, most discriminant selected probabilistic approach named Euclidean Norm Geometric Mean Maximization along Conditional Entropy. Third, performed Canonical Correlation Analysis, aggregated subjected to various classifiers final recognition. evaluation scheme publicly available image dataset CASIA B. We demonstrate methodology, once used Machine, achieves beyond 96%, outperforms existing works several widely adopted classifiers.