This paper introduces a knowledge-driven approach to real-time, continuous activity recognition based on multisensor data streams in smart homes The approach goes beyond the traditional data-centric methods for activity recognition in three ways First, it makes extensive use of domain knowledge in the life cycle of activity recognition Second, it uses ontologies for explicit context and activity modeling and representation Third and finally, it exploits semantic reasoning and classification for activity inferencing, thus enabling both coarse-grained and fine-grained activity recognition In this paper, we analyze the characteristics of smart homes and Activities of Daily Living (ADL) upon which we built both context and ADL ontologies We present a generic system architecture for the proposed knowledge-driven approach and describe the underlying ontology-based recognition process Special emphasis is placed on semantic subsumption reasoning algorithms for activity recognition The proposed approach has been implemented in a function-rich software system, which was deployed in a smart home research laboratory We evaluated the proposed approach and the developed system through extensive experiments involving a number of various ADL use scenarios An average activity recognition rate of 9444 percent was achieved and the average recognition runtime per recognition operation was measured as 25 seconds

A Knowledge-Driven Approach to Activity Recognition in Smart Homes

Ambient intelligence is a new information paradigm, where people are empowered through a digital environment that is “aware” of their presence and context and is sensitive, adaptive, and responsive to their needs. Hence, one of the important requirements for ambient intelligent environments (AIEs) is the ability to localize the whereabouts of the user in the AIE to address her/his needs. In order to protect user privacy, the use of cameras is not desirable in AIEs, and hence, there is a need to rely on nonintrusive sensors. There are various localization means that are available for outdoor spaces such as those which rely on satellite signals triangulation. However, these outdoor localization means cannot be used in indoor environments. The majority of nonintrusive and noncamera-based indoor localization systems require the installation of extra hardware such as ultrasound emitters/antennas, radio-frequency identification (RFID) antennas, etc. In this paper, we propose a novel indoor localization system that is based on WiFi signals which are free to receive, and they are available in abundance in the majority of domestic spaces. However, free WiFi signals are noisy and uncertain, and their strengths and availability are continuously changing. Hence, we present a fuzzy logic-based system which employs free available WiFi signals to localize a given user in AIEs. The proposed system receives WiFi signals from a large number of existing WiFi access points (up to 170 access points), where no prior knowledge of the access points locations and the environment is required. The system employs an incremental lifelong learning approach to adjust its behavior to the varying and changing WiFi signals to provide a zero-cost localization system which can provide high accuracy in real-world living spaces. We have compared our system in both simulated and real environments with other relevant techniques in the literature, and we have found that our system outperforms the other systems in the offline learning process, whereas our system was the only system which is capable of performing online learning and adaptation. The proposed system was tested in real-world spaces from a living lab intelligent apartment (iSpace) to a town center apartment to a block of offices. In all these experiments, our system has been highly accurate in detecting the user in the given AIEs, and the system was able to adapt its behavior to changes in the AIE or the WiFi signals. We envisage that the proposed system will play an important role in AIEs, especially for privacy concerned situations like elderly care scenarios.

A Fuzzy Logic-Based System for Indoor Localization Using WiFi in Ambient Intelligent Environments

Combining ontological and temporal formalisms for composite activity modelling and recognition in smart homes

The elderly population worldwide has an increasing expectation of well-being and life expectancy. The monitoring of elderly people on an individual basis, in a medical sense, will not be a viable proposition in the future. The infrastructure available is not adequate to meet all expectations and subsequently people will continue to live at home with inadequate care. Prior research has shown an accelerated need for the expansion in the Ambient Intelligence (AmI) domain and to that end we present a novel learning technique for intelligent agents that are embedded in Ambient Intelligent Environments (AIEs). A novel agent that combines an emotion recognition system with a fuzzy logic based learning and adaptation technique provides for an automated self-learning system that constantly adapts to individual requirements. This agent, entitled Health Adaptive Emotion Fuzzy Agent (HAOEFA), has the ability to model and learn the user behaviour in order to control the environment on their behalf with respect to his/her emotional preferences. In addition, the agent incorporates temporal adaption in order to facilitate changing behaviour and preferences within the environment. The results show that such architecture can both provide monitoring and ambient environmental control features such that users with limited physical or cognitive functions can have their well-being advanced with limited external resources.

A novel adaptive approach for home care ambient intelligent environments with an emotion-aware system

Activity modelling and discovery plays a critical role in smart home based assisted living. Existing approaches to pattern recognition using data-intensive analysis suffers from various drawbacks. To address these shortcomings, this paper introduces a novel ontology-based approach to activity modelling, activity discovery and evolution. In this approach, activity modelling is undertaken through ontological engineering by leveraging domain knowledge and heuristics. The generated activity models evolve from the initial “seed” activity models through continuous activity discovery and learning. Activity discovery is performed through ontological reasoning. The paper describes the approach in the context of smart home with special emphases placed on activity discovery algorithms and evolution mechanism. The approach has been implemented in a feature-rich assistive living system in which new daily activities can be detected and further used to evolve the underlying activity models.

A systematic approach to adaptive activity modeling and discovery in smart homes

https://victor.callaghan.info/publications/2009_pmc09%28MakingOurEnvironmentsIntelligent%29.pdf

Abdesalam Helal

Papers

Editorial: Making our environments intelligent