Achieving Pro-Active Guidance of Patients through ADL using Knowledge-Driven Activity Recognition and Complex Semantic Workflows

Assisted Ambient Living (AAL) focuses on self-sufficiency, assisting disabled people (in particular, cognitive decline) to perform activities of daily living (ADL) such as housework and taking medication, by automating assistive actions in smart environments. We argue that AAL provides opportunities for pro-active assistance of cognitively disabled patients, which involves dynamically guiding them through an ADL and correcting their actions when required. Activity recognition is a pivotal task in this effort, since it allows detecting when an ADL is started by recognizing its constituent activities. When dealing with diseases such as cognitive decline, activity recognition should be able to detect when activities are performed incorrectly as well – e.g., performed out-of-order, at the wrong location or time, or with the wrong objects (e.g., utensils) – which is nevertheless not a common goal in knowledge-driven activity recognition. In this paper, we present an approach to computerize complex ADL workflows, using an OWL ontology to represent tasks and their temporal relations, in order to realize continuous, pro-active patient assistance. This process is supported by fine-grained, knowledge-driven activity recognition, which employs semantic reasoning to recognize both correct and incorrect actions based on their associated context and temporal relations.