A Novel GAIT Classification Approach Using ELM

Analyzing human gait has earned considerable interest among Computer Vision Community researchers as it has immense use in deducing the physical well-being of people. In this paper, a novel machine learning approach Extreme Learning Machine (ELM) normalized with T-Test is used to detect unusual gait patterns. Extreme Learning Machine classifiers are powerful tools, specifically designed to solve large-scale classification problems. In ELM, one may randomly choose and fix all the hidden node parameters and then analytically determine the output weights of Single-hidden Layer Feed forward neural Networks (SLFNs). After the hidden node parameters are chosen randomly, SLFN can be considered as a linear system and the output weights can be analytically determined through a generalized inverse operation of the hidden layer output matrices. ELM avoids problems like local minima, improper learning rate and over fitting which are commonly faced by the previous iterative learning methods. It also completes the training very fast. The multi category classification performance of ELM with T-Test and PCA are evaluated with Virginia Gait database. The results indicate that ELM produces better classification accuracy while reducing the system complexity and the training time.