Ontology‐based activity recognition in intelligent pervasive environments

Abstract

Purpose -This paper aims to serve two main purposes. In the first instance it aims to it provide an overview addressing the state-of-the-art in the area of activity recognition, in particular, in the area of object-based activity recognition. This will provide the necessary material to inform relevant research communities of the latest developments in this area in addition to providing a reference for researchers and system developers who ware working towards the design and development of activity-based context aware applications. In the second instance this paper introduces a novel approach to activity recognition based on the use of ontological modeling, representation and reasoning, aiming to consolidate and improve existing approaches in terms of scalability, applicability and easy-of-use.Design/methodology/approach -The paper initially reviews the existing approaches and algorithms which have been used for activity recognition in a number of related areas. Form each of these their strengths and weaknesses are discussed with particular emphasis being placed on the application domain of sensor enabled intelligent pervasive environments. Based on an analysis of existing solutions, the paper then proposes an integrated ontology-based approach to activity recognition. The proposed approach adopts ontologies for modeling sensors, objects and activities, and exploits logical semantic reasoning for the purposes of activity recognition. This enables incremental progressive activity recognition at both coarse-grained and fine-grained levels. The approach has been considered within the realms of a real world activity recognition scenario in the context of assisted living within Smart Home environments.Findings -Existing activity recognition methods are mainly based on probabilistic reasoning, which inherently suffer from a number of limitations such as ad-hoc static models, data scarcity and scalability. Analysis of the state-of-the-art has helped to identify a major gap between existing approaches and the need for novel recognition approaches posed by the emerging multimodal sensor technologies and context-aware personalised activity-based applications in intelligent pervasive environments. The proposed ontology based approach to activity recognition is believed to be the first of its kind which provides an integrated framework based on the 2 Liming Chen, Chris Nugent unified conceptual backbone, i.e., activity ontologies, addressing the lifecycle of activity recognition. The approach allows easy incorporation of domain knowledge and machine understandability which facilitates interoperability, reusability and intelligent processing at a higher level of automation.Research limitations/implications -There are a number of challenges which have not been addressed within this paper. These mainly relate to the explicit modelling of the temporal dimension of the problem and managing uncertainty in both the sensor and decision making process. These will be key areas for consideration within our future work.Practical...

Bridging e-Health and the Internet of Things: The SPHERE Project

“…Among existing solutions, physiological, environmental and vision sensors are frequently used for the purpose of health monitoring in the home. These help to characterise users' everyday activities for dedicated purposes by providing long-term sensing data that, in combination with ambient intelligence algorithms, contribute to behaviour pattern recognition [1]. The following subsections reflect the distinction that is frequently adopted in the literature, owing to the different data outputs and approaches required for each.…”

Dynamic sensor data segmentation for real-time knowledge-driven activity recognition

“…Whenever sensor data are obtained, the approach determines the likely ADL by reasoning against the model through ontological inference. Nevertheless, existing works on ontology-based activity recognition [8,9,20] and similar work on knowledge-driven activity recognition [16,17,21] do not clearly articulate the mechanism about how and what sensor data are selected from a live data stream for performing activity inference. In some research experiments that support on-line continuous activity recognition, the experiments restart manually each time an ADL is identified.…”

OWL 2 modeling and reasoning with complex human activities

“…For this reason, the use of formal ontologies, specified using the OWL 1 language or its ancestor DAML+OIL, has been investigated to represent context data, from raw ones acquired from sensors, to complex ones such as human activities (e.g., the ontologies SOUPA [20], CONON [21], the one used in CARE [22], and the one for smart homes used in [23]). In particular, a technique to recognize human activities based on ontological reasoning alone has been proposed by L. Chen et al [7,23]. The main idea is to use ontologies not only to represent activities, but also each data that can be used to recognize them, including sensors, objects, locations, and actors.…”

“…These are the logics underlying the popular ontological language OWL 1 that has also been used to build activity ontologies in the area of pervasive computing [7,8]. The ontological approach to activity modeling consists in a knowledge engineering task to define the formal semantics of human activities by means of the operators of the ontological language.…”