A new fast and calibration-free method for footstep impact localization in an instrumented floor

Abstract

Impact localization in a floor is complicated due to the dispersion-caused distortion of the generated floor waves. Current localization methods that try to overcome the dispersion problem are computationally expensive, taking in some cases 2 seconds to yield a single footstep location estimate. If an accelerometer sensor network is utilized to localize footsteps, and consequently track an occupant's path, then there is a need for computationally fast algorithms that are able to keep up with the walking (or running) impact frequency; therefore, in this paper, a practical algorithm is proposed for footstep impact localization in an instrumented floor. The proposed algorithm has promising sub-meter localization accuracy and is computationally fast. In addition, the algorithm does not require estimation of floor-dependent parameters, which is an additional advantage since estimating floor-dependent parameters in a floor will have relatively high uncertainty as the floor cannot be treated as an isotropic/homogeneous material. The proposed algorithm is evaluated using simulations and an experiment in an operational smart building.

A passive energy-based method for footstep impact localization, using an underfloor accelerometer sensor network with Kalman filtering

“…Previous energy-based localization methods in dispersive waveguides include the works in Alajlouni et al (2017, 2018) and Alajlouni and Tarazaga (2019a, 2019b). In Alajlouni et al (2017), an energy-based impact localization method is evaluated utilizing the response of a simulated finite element plate model to virtual hammer impacts.…”

“…Then, the study in Alajlouni and Tarazaga (2019a) presented the first heuristic energy-based localization method to be tested with the footsteps of a walking occupant. The method estimates a footstep’s location as a weighted sum of sensor coordinate vectors, where the weights are proportional to the average power, Q , calculated at each sensor.…”

“…The method estimates a footstep’s location as a weighted sum of sensor coordinate vectors, where the weights are proportional to the average power, Q , calculated at each sensor. The heuristic method in Alajlouni and Tarazaga (2019a) is calibration-free; that is, it does not require estimation of a floor-dependent parameter. In addition, the method is computationally cheap.…”

“…In this paper, the authors present a new localization algorithm that modifies the method in Alajlouni et al (2018) by estimating, for each detected footstep, a suitable value of β online rather than estimating a fixed value of β offline during a calibration phase, thus making the new method calibration-free. Moreover, instead of using a fixed initial search point to seed the nonlinear minimization solver as is the case in Alajlouni et al (2018), the new algorithm utilizes the heuristic location estimate in Alajlouni and Tarazaga (2019a) to seed the nonlinear minimization solver.…”

CapLoc: Capacitive Sensing Floor for Device-Free Localization and Fall Detection

“…Floor-based localization can also be accomplished by measuring footsteps-induced vibrations with a network of seismic sensors [23][24][25]. The footsteps (and hence the target) are located by exploiting the fact that vibration signals take different times to reach each sensor depending on the distance between the footstep and the sensors.…”