In a general aspect, a method is presented for detecting a location of motion using wireless signals that propagate along two or more paths of a wireless communication channel. The method includes storing a set of eigenvectors derived from first motion-sensing data associated with a first time frame. The first motion-sensing data is associated with a first motion-sensing topology of a wireless mesh network. The method also includes obtaining a motion vector based on wireless signals transmitted between access point nodes in the wireless mesh network during a second, subsequent time frame. The wireless mesh network operates in a second, distinct motion-sensing topology during the second time frame. The motion vector is compared with the respective eigen vectors, and a probability vector is generated based on the comparison. A location of the motion of the object during the second time frame is determined based on the probability vector.