Human Gait Analysis with Hybrid Deep Learning Network CNN-SVM

In the world of digitization, Security concerns arise in private as well as public places like airports, metro stations, banks, and even at homes. This has led to diversity in the research area, and modern video surveillance systems can be considered one of them. One of its’ subsets can be put under Human Gait, where brilliant minds have already provided a breakthrough in the research industry. Human Gait involves the movement of limbs during locomotion it comprises of various ways in which a human being can walk. Now coming to human gait recognition, these systems can recognize the walking style of a person from a pretty long distance despite of any interaction with the machines as in biometrics. In this paper, the implementation follows Human Gait Recognition using Feature extraction through Convolutional Neural Networks (CNN) and Classification through Support Vector Machine (SVM). The preferred framework involves Data pre-processing, Feature Extraction, and Classification. The CNN SVM was trained for 9041 images out of which some were kept for testing and some for-testing purposes. The model was trained for the first four steps of the human gait cycle including Initial Contact, Loading Response, Mid Stance, and Terminal stance. Recognition of these four walking steps along with their accuracy rate. The maximum output proved out to be for Mid Stance which is 98.78%. The future scope of this implementation is much more awaited and can help in recognizing a person by identifying the walking patterns of an individual. This study may find its way in all IoT devices which will help in recognizing and authenticating an individual.