Lots of aging people rely on a cane as a help device to ovoid many problems, such as balance loss and leg weakness, presumed to be at the origin of many fall incidents. In this paper, we present a new design and implementation of Canne3D system, a local monitoring system for elderly using their connected walking stick. Actually, the 3Dcane is modeled through consideration of real-time movements for a thorough understanding of the concerned person's state, including fall detection, steps counting and calculating the distance traveled. We have applied a multi-stage thresholding framework based on the acceleration value. Experiments have revealed that step-detection algorithm appears to be robust in term of detection rate and false positive rate.