Accidental falling is one of the health hazards for seniors. A well falling detection mechanism can help the caregivers to save the elders in time. Falling detection technique becomes a critical factor to improve the quality of home care. Generally, the acceleration of a human's waist is smooth. Only the unstable state will generate acute acceleration. Therefore, in this paper, we installed the G-sensor on waist. We used the returned values of two installed G-sensors and the slopes of their signal measure vector to detect whether the tested man is falling or not in the sitting on the chair and standing cases. Our experimental results show that the falling can be detected when the slope of the signal measure vector rises steeply.