A New Method of Gait Recognition by Human Joint Projection and Closest Distance Technique

Abstract

Smart home is an alternative function in present accommodation. Many technologies are developed for smart home. Human gait recognition system for indoor area is applied for monitoring purpose such as elder health care system. However, simple and effective algorithm is necessary for real-world applications. In this paper, we propose a new method of human gait identification system for indoor area by using human joints projection and closest distance technique. The algorithm consists of four step: image sequence acquisition, human joints segmentation, human joints projection and gait recognition. Firstly, human 2D sequence image is acquired. Red nine markers are attached on human joints. Secondly, markers are extracted by color thresholding technique. Human joints are then tracked. Third, such joints is projected in a dimension. Finally, Euclidean distance technique is used to compare two 2D human points between test data and train data. Human gait is recognized by the closest distance. To evaluate the proposed method, fifteen subjects walk pass 2D camera in same one way. A subject was tested three time. Average accuracy rate is 95.55%. This method perform effectively. The advantages of this method over existing methods is simple algorithm and effective for small size of target as using in accommodation. © 2017 IEEE.