An Efficient Algorithm to Acquire Displacement from Live Streaming Acceleration

Computing displacement from measured acceleration is important for various computer vision activities. Such displacements are not obtained easily with a good accuracy. The aim of this paper is to develop an algorithm to acquire the displacement realized from a mobile platform such as a wheelchair travelling a definite distance. The algorithm uses mainly the information from the accelerometer and minimizes the error between the desired displacement and the actual displacement travelled by the mobile platform. In essence, this paper proposes a novel technique to derive displacements using the Fourier transformed acceleration divided by the scaling factor of -ω2 while zero-padding the noise signal to reduce the margin of errors and their size. Experiments prove that the system is reliable and provides accurate pose estimates. The proposed algorithm was tested on real indoor environments using Matlab software and the experimental results demonstrate its precision compared to other techniques.