On-line learning based on spiking neurons for human state estimation in informationally structured space

Abstract

Recently, as the number of elderly people increases, much more caregivers are required for the support for them. However the number of caregivers and therapists is not enough in the current situation. In this paper, we propose a support system for the elderly introducing robot partners, sensor networks, and portable sensing devices in informationally structured space. In the system, human state estimation is one of the most important technologies. In order to realize the estimation suitable to the elderly, we should consider how to model the human states. Most of previous methods are based on off-line statistic approaches. In this paper, we discuss an on-line learning method for modeling human states. First of all, we explain the system for the elderly in informationally structured space. Next, we propose an on-line learning architecture based on spiking neurons. Finally, we show an example of experimental result for modeling the patterns of human states in a living room.