Human Gait Recognition

Abstract

As the security awareness is increasing, individuals need to utilize the more helpful distinguishing proof strategies on human recognition - biometrics responses to this interest. Gait Recognition is a promising technique in biometric innovation. The method recognizes user/ individuals dependent on their walk style. For many years, gait recognition systems are underdeveloped because of very less attention towards gait analysis. Only a few researches have been carried out on the subject and is often considered a highly advanced computer vision problem. Other systems like face recognition, action recognition have been extensively developed and very less efforts in developing efficient gait analysis system. Most of the researches are carried out on basic machine learning algorithm which are prone to bias problem and often feature reduction has to be done in order to overcome bias. Deep learning provides an all-in-one solution automatically and once the algorithm is trained it automatically learns all the features to classify person with his gait analysis. Past methods for example fingerprints, retina, palm and voice recognition were used to identify individuals. Still, it needs people’s permission and physical attention. However, Human Gait Recognition takes a shot at the step of strolling subjects to distinguish individuals without them knowing or without their consent. The motivation behind this report is to sum up our examination and related work on Human Gait Recognition and algorithm which assists in distinguish strolling subject from a distance without any authorization and obstruction of the walking item. We have tried to achieve all this by making a database of the recorded videos from a camera (converted them into the frames of still images) and images of individuals in the eastern dresses. Later, we applied extraction techniques to obtain the silhouettes. Silhouettes then trained with principal component analysis. Then, the pictures in the database were matched with the input images. If both the images match, the system recognizes or identifies the individual or else save the new image in the database. In this project we have driven deep into all the research carried out in gait analysis and by using deep learning tried to resolve said computer vision problem.