Using Fuzzy Logic to Enhance Classification of Human Motion Primitives Barbara Bruno 1 , Fulvio Mastrogiovanni 1 , Alessandro Saffiotti 2 , and Antonio Sgorbissa 1 1 University of Genova, Dept. DIBRIS, via Opera Pia 13, 16145 Genova, Italy {barbara.bruno,fulvio.mastrogiovanni,antonio.sgorbissa}@unige.it 2 ¨ Orebro University, AASS Cognitive Robotic Systems Lab., Fakultetsgatan 1, S-70182 ¨ Orebro, Sweden asaffio@aass.oru.se Abstract. The design of automated systems for the recognition of spe- cific human activities is among the most promising research activities in Ambient Intelligence. The literature suggests the adoption of wearable devices, relying on acceleration information to model the activities of interest and distance metrics for the comparison of such models with the run-time data. Most current solutions do not explicitly model the uncer- tainty associated with the recognition, but rely on crisp thresholds and comparisons which introduce brittleness and inaccuracy in the system. We propose a framework for the recognition of simple activities in which recognition uncertainty is modelled using possibility distributions. We show that reasoning about this explicitly modelled uncertainty leads to a system with enhanced recognition accuracy and precision. Keywords: Activity recognition, Activities of Daily Living, wearable sensors, possibility measures. 1 Introduction Automatic recognition of human activities is a vivid area of research, whose impact ranges from smart homes to future factory automation and to social behavioural studies. One of the most timely application is the process of deter- mining the level of autonomy of an elderly person. Ever since the publication of the Index of Activities of Daily Living (ADL) by Katz and colleagues [1], this process is usually accomplished by analysing the person’s ability to carry out a set of daily activities, each one involving the use of different motor and cogni- tive capabilities. Unfortunately, the most commonly adopted indexes and sets of ADL have been defined assuming that a caregiver examines the person’s per- formance on a qualitative basis; this makes the design of automated systems for the monitoring, recognition and classification of ADL particularly challenging. Existing solutions take two different approaches: smart environments rely on heterogeneous sensors distributed in the environment [2–4]; wearable sensing systems rely on sensors located on the person body [5, 6]. A. Laurent et al. (Eds.): IPMU 2014, Part II, CCIS 443, pp. 596–605, 2014. c  Springer International Publishing Switzerland 2014