The invention discloses a falling detection method, and belongs to the fields of behavior analysis and identification methods. The falling detection method comprises the following steps of S1, enabling a sensor to collect accelerations of a human body in X-axis, Y-axis and Z-axis directions, calculating a resultant acceleration a of the human body corresponding to a point of time, and making a real-time coordinate map of the relationship between time t and the resultant acceleration a; S2, setting a standard coordinate map, wherein the standard coordinate map is a coordinate map which is made by analyzing the relationship betwen time T and resultant acceleration A when the human body falls down; S3, setting an impact marker, and executing the next step when the impact marker is triggered; S4, analyzing and processing the real-time coordinate map, comparing with the standard coordinate map, calculating the size of an amaz/Amzx value to distinguish whether the falling occurs or not, and judging a falling type and a falling degree. The falling detection method aims to achieve the advantages that the accuracy of detection on falling of the human body is improved, a human body falling detection algorithm is simplified, and a serious consequence of a detected person due to falling is greatly decreased.