0 avis
Human sensing in reverberant environments: RF-based occupancy and fall detection in ships
Archive ouverte : Article de revue
Edité par HAL CCSD ; Institute of Electrical and Electronics Engineers
International audience. In this paper, we explore the possibility of estimating the number of people inside a below-deck ship compartment using the theory of room electromagnetics. Most RF-based occupancy detection solutions focus on residential environments. Confined metallic spaces such in industrial environments are more hostile and exhibit features of a reverberant cavity. The theory of room electromagnetics provides a simple characterization of microwave propagation in such environments. By considering the indoor environment as a lossy cavity, the exponential decay rate of the power-delay profile (PDP), also known as reverberation time (RT), is related to the total absorption inside the room. First, we verify the reverberating nature of the room by measuring the RT at different locations inside the room. The PDPs are calculated from the channel impulse response (CIR) measured using the MIMOSA radio channel sounder. Then, the relation between RT and the number of people inside the room is investigated. We show that it is possible to estimate the number of stationary people with a good accuracy, depending on the number of antennas used. With a success rate of 88%, the estimation error is only one person when 16 spatially averaged PDPs are used. Higher success rates are achievable with more spatial averaging. Moreover, off-the-shelf (OTS) ultra wide band (UWB) devices are used to estimate the number of people inside the same room. Results show that temporal averaging of PDPs can further improve the success rate to 95% when people inside the room are moving. In addition, the detection of a lone person falling to the ground is investigated based on the Doppler analysis of the measured CIRs. It is shown that the Doppler spread in case of a fall has higher peak values compared to normal activities, such as walking and sitting. With the use of a simple Bayes classifier, a fall is detected with 98% accuracy and 100% sensitivity.
Consulter en ligne
Suggestions
Du même auteur
Du même sujet
