Human step detection from a piezoelectric polymer floor sensor using normalization algorithms

Today, walking trajectory analysis or event detection gained a lot of interest in healthcare environments to prevent pathologies or gait deviations. However, to correctly study the gait of someone is difficult since it generally requires expensive systems that are difficult to implement. In the work presented herein, we propose another approach to retrieve information from a sensor area. As usual gait analysis systems comprise numerous sensors and provide spatial pressure resolution maps of footsteps over a small area; our system gives a signature generated by a single piezoelectric sensor only. The signal delivered by the sensor goes through a normalization calculation process giving a signal which can be compared to a reference signal. A Pearson product-moment correlation coefficient (PPMCC) is determined between the two signals giving a shape similarity indicator. We achieved similarity indicators greater than 95%. It is the first time that this correlation is applied in the field of human step detection.