Evaluating the recognition of bed postures using mutual capacitance sensing

Article type: Research Article

Authors: Rus, Silvia^{a; *} | Grosse-Puppendahl, Tobias^b | Kuijper, Arjan^{a; c}

Affiliations: [a] Fraunhofer IGD, Fraunhoferstr. 5, 64283 Darmstadt, Germany. E-mails: [email protected], [email protected] | [b] Microsoft Research, 21 Station Road, CB1 2FB Cambridge, United Kingdom. E-mail: [email protected] | [c] Technische Universität Darmstadt, Hochschulstr. 10, 64289 Darmstadt, Germany. E-mail: [email protected]

Correspondence: [*] Corresponding author. E-mail: [email protected].

Abstract: Capacitive sensing is increasingly used to gather contextual information about humans. They can be used to create stationary or mobile systems for non-contact activity recognition. They are able to sense any conductive objects at distances up to 50 cm. This paper investigates an approach to classify bed postures using mutual capacitance sensing. The goal is to develop a system that prevents decubitus ulcers, which is a condition caused by prolonged pressure on the skin that can result in injuries to the skin and underlying tissues. The posture recognition is used to detect prolonged lying in a single pose and can notify care personnel. A low-cost grid of crossed wires is proposed that is placed between the mattress and the bed sheet that creates 48 measurement points. The experiments analyze a set of five bedding positions with 14 users. Using self-defined features, we achieved an accuracy of 80.8% for all users and an accuracy of 93.8% for individuals of similar body size. Refining the classification approach by directly classifying the raw data an overall accuracy of 90.5% was reached. By introducing an uncertainty threshold the classification is correct in 97.6% of cases.

Keywords: Activity recognition, capacitive sensing, Ambient Assisted Living

DOI: 10.3233/AIS-160414

Journal: Journal of Ambient Intelligence and Smart Environments, vol. 9, no. 1, pp. 113-127, 2017