ABSTRACT
The rapid increase in the population of elderly people will demand the development of new analytic tools since elderly people suffer from the consequences of cognitive impairment. This affects the cognitive abilities of elderly people and causes problems with learning and memory, which makes it difficult to have an independent life for elderly people. If the indicators of cognitive impairment could be detected at an early stage, early treatment and further diagnosis could be applied. Thus, in this chapter, a method is presented to track the daily life activities of elderly people at a smart-home and it detects the abnormal behaviour, which might arise from the consequences of cognitive decline. This chapter presents a three-step methodology for this purpose: (1) A method is proposed to simulate the abnormal behaviour of elderly people suffering from cognitive decline since it is difficult to collect real-world data. (2) Deep learning methods, namely Recurrent Neural Networks (RNNs) and Convolutional Neural Networks (CNNs) are adapted to recognise daily life activities and then detect abnormal behaviour. (3) The results are compared to the state-of-the-art methods such as Conditional Random Fields (CRFs) and Hidden Markov Models (HMMs). The results show that RNNs and CNNs are promising methods to support the decision-making process of medical doctors and caregivers to detect the early indicators of cognitive decline.
