Identification of gait data using machine learning technique to categories human locomotion

Abstract

Human Gait is a model of movement while locomotion and it is known as behavioural characteristic of human. Gait recognition is a behavioural biometric method that includes human identification from afar i.e. in absence of human and this is achieved by analysing the behaviour of their walking. Main aim to make GAIT recognition system is to impart a best method to recognize risks in places where high security is required like in banking, parking, airports etc. or to detect disease like Parkinson's. In this paper we have shown the importance of gait recognition in order to detect whether a human GAIT is normal or abnormal, firstly the features of human gait is collected and then they are classified using neural network (Back propagation) and KNN classification technique. To test the proposed system the database contains 204 gaits and 16 different features were recognised, in which 3 datasets are crouch gait dataset and one normal gait dataset of 4 different humans is collected. We have obtained 100% of classification accuracy for a training data. Testing results shows 39.65% accuracy overall accuracy of system is 69.825%.