Human action estimation based on non-dedicated sensor information for environmental control in university library

Due to global warming, energy management systems (EMSs) for spaces are attracting a lot of attention. In order to realize effective EMS control and thus higher energy efficiency, we must be able to detect human presence/action in each space. Using non-dedicated sensor information to achieve this can reduce the number of sensors and the network workload as well. This paper proposes a simple method for estimating human presence and action as well as a new signal processing procedure that uses basic non-dedicated sensor information, the received RF signal level and its deviation. This method is simple and has the exciting benefit of compatibility with existing devices and existing wireless sensor networks. Two experiments verify the performance of the proposed method. A provisional mask in terms of received signal level and deviation area is produced in a pre-experiment in a laboratory building. Then, in an actual field trial in a university library, the mask is refined and estimation accuracy is evaluated. Estimation accuracy is 75 % for standing status, 83 % for sitting status, and 76 % for walking status in the space between bookshelves. All are expected to improve with the use of multiple paths.