Inference Engine Based on a Hierarchical Structure for Detecting Everyday Activities Within the Home

One of the key objectives of an ambient assisted living environment is to enable elderly people to lead a healthy and independent life. These assisted environments have the capability to capture and infer activities performed by individuals, which can be useful for providing assistance and tracking functional decline among the elderly community. This paper presents an activity recognition engine based on a hierarchical structure, which allows modelling, representation and recognition of Activities of Daily Life (ADLs), their associated tasks, objects, relationships and dependencies. The structure of this contextual information plays a vital role in conducting accurate ADL recognition. The recognition performance of the inference engine has been validated with a series of experiments based on object usage data collected within the home environment.