This thesis implements the elderly fall detection which is based on human shape and shadow elimination technique. The human shape mechanism includes standing, bending, squatting and lying down postures. The shadow elimination technique solves judged mistakes caused by overlapping shadow generations in an inefficient light environment. Therefore, according to the human shape mechanism, the system analyzes the behavior of elderly activities in the daily life. Then, the shadow elimination technique can be together with the human shape mechanism for feature extractions in height, distance and speed parameters of data. Data results can accurately judge the posture and falling detection for elderly actions in the daily life.