Recognition of human activities by gas sensor response using genetic algorithm

Abstract

AbstractIn this paper, a scheme for estimating gas concentration change signals for each gas generating source in a room from output signals of a combustible gas sensor installed in the room, and for estimating event signals synthesized from them by a genetic algorithm (GA), is proposed. It is shown by computer simulation that more accurately estimated concentration change signals or event signals are obtained by applying selective weeding, crossing, mutation, and so on, repeatedly to an individual group, while applying an appropriate evaluation function to concentration change signals and event signals estimated for each gas generating source expressed by a GA individual. In addition, it is shown that the estimated concentration change signals and estimated event signals obtained from the output signals of a gas sensor installed in an occupied room by the proposed scheme correspond approximately to human activities. In addition, it is shown that the proposed scheme is effective in detecting human activities with a gas sensor, since appropriate signals can be estimated robustly even though components such as noise may be present in the sensor output signals. © 2001 Scripta Technica, Electron Comm Jpn Pt 3, 85(1): 50–61, 2002