In recent years, the service robotics market has been continuously expanding. A service robot needs to move and work among many people; therefore, it should be able to avoid obstacles and unnecessary people and provide service to specific people. Following robots, which assist people in carrying luggage at airports and hotels and packages at factories, have been developed. A following robot detects a person using a sensor and continues to follow him/her, but often loses sight of that person after a while. In this study, we attempt to detect a person using the Xtion® depth sensor, manufactured by ASUS. We propose a way to identify people using "shoulder width" and "head height" in a triangle consisting of three points on the head, left shoulder, and right shoulder. The parameters are calculated from the data obtained from the face and torso.