Abstract
Ontology-based knowledge driven Activity Recognition (AR) models play a vital role in realm of Internet of Things (IoTs). However, these models suffer the shortcomings of static nature, inability of self-evolution and lack of adaptivity. Also, AR models cannot be made comprehensive enough to cater all the activities and smart home inhabitants may not be restricted to only those activities contained in AR model. So, AR models may not rightly recognize or infer new activities. In this paper, a framework has been proposed for dynamically capturing the new knowledge from activity patterns to evolve behavioural changes in AR model (i.e. ontology based model). This ontology based framework adapts by learning the specialized and extended activities from existing user-performed activity patterns. Moreover, it can identify new activity patterns previously unknown in AR model, adapt the new properties in existing activity models and enrich ontology model by capturing change representation to enrich ontology model. The proposed framework has been evaluated comprehensively over the metrics of accuracy, statistical heuristics and Kappa Coefficient. A well-known dataset named DAMSH has been used for having an empirical insight to the effectiveness of proposed framework that shows a significant level of accuracy for AR models This paper is a postprint of a paper submitted to and accepted for publication in IET Wireless Sensor Systems and is subject to Institution of Engineering and Technology Copyright. The copy of record is available at the IET Digital Library
Highlights
Human Activity Recognition (HAR) determines the activities that have been performed by humans based upon certain knowledge and context
N and et al [13][33] compare the database schema evolution with ontology evolution and conclude that ontology evolution is similar to object-oriented database schema and semantically richer than database schemas due to it inheritance principles and conclude that ontology evolution approaches are a kind of extension rather than an adaptation of existing approaches
In order to evaluate the performance of the proposed framework and have an empirical view of its effectiveness, baseline datasets described in [1] named Data Acquisition Methodology for Smart Homes (DAMSH) has been used
Summary
Human Activity Recognition (HAR) determines the activities that have been performed by humans based upon certain knowledge and context. We have already developed a seed ontology described in [7] specifying a generic set of actions for each activity called perceptible activity model (PAMs) from which personalized/complete set of actions (i.e. CAMs) for particular activities are derived These personalized patterns along with their labels are stored in a log file. The proposed framework is envisaged to have the following features: (i) Ability to learn new activities by exploiting the context of existing activities and activity patterns in log file It may learn the specialized activity of existing activity or a new activity from unidentified patterns (ii) Ability to reuse the existing knowledge of ontology to learn an activity Enable the model to be adaptable and flexible. Adaptability and flexibility enable the model to dynamically update its existing activity model like an action sequence, duration etc
Sign-in/Register to view highlights & summary Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
