A Multimodal Human Sensing System for Assisted Living

Abstract

Introduction: Advancements in sensor technology have resulted in smart systems that can analyze, interpret and understand their surroundings for decision making. Using such intelligent systems for environments such as smart homes, device automation, and hospitals, makes sensing human presence an essential requirement. Human sensing becomes critical, especially in assisted living scenarios such as in elderly care. However, mechanisms for human sensing are not foolproof because of the dynamic nature and ability of human beings to deliberately mislead the sensors.

Objective: Objective of paper is to detect human presence in varying environment using non-intrusive sensors.

Method: Besides, sensors have inherent limitations, due to either their mechanism of sensing or the environmental conditions, which can cause them to fail in human detection. These limitations can, at times, cause sensors to provide useless data to the system. In this paper, we propose an adaptive multi-modal human sensing mechanism which can autonomously identify and ignore unnecessary data from a set of sensors, thereby reducing computation complexity, reducing false alarm rate and yielding better performance. The effect of sensing when the human being is in motion has also been studied.

Results: The results portrayed in the paper prove the efficacy of the proposed multi-modal system over itssingle sensor counterparts when used in changing environments and other proposed multi-modal human sensing.

Conclusions: Human sensing is vital to many smart applications such as smart homes, traffic management systems, human-computer interfaces, etc. Since human beings, in general, are always on the move, the use of a dedicated sensor could fail to detect human presence, especially when the ambient parameters around the sensor change. In this paper, a multi-modal human sensing approach has thus been prescribed for overcoming this issue. This work described focused on automating the identification of inappropriate data relayed by some sensors under certain environmental conditions. Experiments reported include both cases - when the sensors are mounted on a static unit (door frame) and also on a mobile robot. The corresponding results reveal that a combination of sensors outperforms the use of individual dedicated sensors for human detection. Analyses of the walking speed of human being have also been studied which endorse the robustness of the approach. However, different directions of motion need exploration in future.