A capacitive proximity device SensFloor [1] was installed in the HUman at home project [2] appartment in Montpellier, France. Floor activations related to participants' movements are recorded with a frequency of 868Hz. One issue of this low-cost floor is its poor spatial precision. Actually, the least spatial element is a triangle of 25 × 50 cm. Moreover, this capacitive signal is not proportional to the weight of the object. Thus, it is challenging to organize this space-temporal signals into human behavioral events such as static, trample, walk. These events are identified by our Walk@Home algorithm presented here. In the core of this algorithm is a space-temporal window that scans the raw signals and organize them into a dynamic graph containing the eventual trajectories. Even then, the result is an approximation of the movement of the center of the gravity of a human. A good identification of trajectories is validated with data from controlled movements on the floor.