In this study, we describe the analysis of human walking towards the development of biped robot. There are a number of previous studies on the behavior analysis of human. A way to get the skeleton information, an example of using the KINECT developed as a controller for the game available at low cost has increased in recent years. In this study, the examples of motion are normal walking, the left hemiplegia walking, robot walking and walking with operated smartphone. The measurement method, were taken by moving to match the walking speed of a person and the KINECT on the truck. In addition, we analyzed the eigenvalues by the singular value decomposition skeleton information. Result of the analysis were in the robot walking is most stable. As the result, it is possible to analyze feature quantity of the walking is obtained by singular value decomposition. It is believed that the application of this method, it would be utilized for prediction and prevention of falling during walking.