Spatio-Temporal Boolean Compressed Sensing for Human Localization With Fiber-Optic Sensors

Human localization with fiber-optic sensors is a data-efficient, low-computational-cost substitute for vision-based approaches, especially, in indoor environments. A challenging task in building such a system is increasing the sensing efficiency, that is, the ratio of the number of monitored cells to that of sensors involved. In this paper, we develop a spatio-temporal Boolean compressed sensing model for addressing this issue. Specifically, we formulate the sensing task as the issue of encoding and decoding the sensed space in a joint spatio-temporal fashion, and we employ ant colony optimization for creating the required codebook. The design and implementation of this model is presented as well. Two aspects are mainly concerned. First, the modular design paradigm is explored to facilitate the deployment scalability. Second, a calibration mechanism is incorporated into the signal acquisition process for reliability enhancement. A lab-scale prototype system is developed for localizing two persons within a 7 x 7 grid using only 12 sensors, which are more efficient compared with 15 sensors required in a conventional model. The experimental results are reported to validate the proposed model.