Gait Feature Extraction. Parallel Lines Under The Constraint Of The Gait Recognition Algorithm Based On Differential Movement Slice Study

Gait recognition concerns how to get the identification information of the objects recorded in videos. In the pretreatment, Image processing and computer vision techniques are employed to track and separate the objects from the background; in the recognizing process, pattern recognition and artificial intelligence are used to characterize and analyze the walking styles of the objects. Traditional biometrics used for identification include fingerprint, voice, iris, face, signature and so on, out of which fingerprint and face are most widely used. The growing demand for automatic human identification system in many important applications has made the recognition based on gait a hot research point in recent because of its advantages over other biometrics identifications. Gait is observable and difficult to disguise. Especially, it can be recorded by cameras at a distance, which enables it to work as an ideal noninvasive biometrics for identifications in such important implications as public surveillance and security. However, gait recognition is still in the stage of researching. There is still a great distance between this novel biometrics and workable system that can be put into the real world.A great deal of researches on gait recognition have been conducted in recent years, including medical and psychological studies, both abroad and at home. In this work, a deep analysis and classification on the previous researches is made. For the most gait recognition methods, it's impossible to conquer the problems caused by view-point difference and clothing difference. A detailed discussion on these problems is made in the paper, then several gait analysis and recognition methods are proposed to the view-point difference problem.One of the most important aspects in gait recognition is how to character human walking and extract gait signatures from video sequence. According to whether body structure is modeled, the characterization of walking in this work falls into 2 categories, namely model-free and model-based methods. In the model-free category, difference motion slice is proposed, which then is processed to feature the whole walking sequence. In the model-based ones, an inter-linked ellipses model is proposed to characterize the...