FuturAAL-ego: Complex and scalable egocentric vision-based activity recognition for AAL


The problems derived from the ageing of population in developed countries are a serious threat to the sustainability of the welfare state. As ageing related health expenses are continuously growing, it is mandatory to explore new ways towards active and independent ageing. One of the ways to achieve those objectives is to apply information and communication technologies (ICT) to develop automatic assistive systems (the Ambient Assisted Living or AAL paradigm). Such automatic assistive systems are designed to help people performing the so called Activities of Daily Living (ADL), as a person who can accomplish ADLs is said to be independent. In consequence, assistive systems should work on any environment where elderlies perform their daily activities, i.e. in urban environments. However, current systems are usually limited to more controlled environments as homes. The base of an assistive system is to properly understand the situation of an elder, in order to provide the needed assistance. We argue that current systems cannot work in urban or non-structured environments because their situational awareness approach is not suitable for such environments. Thus, this project proposes novel research on situational awareness systems for AAL. First of all, we propose to use wearable cameras for elderlies to avoid any dependency with the environment. Using wearable devices we avoid external infrastructures and we can capture very rich information about the environment, objects and people. Second, we will divide situational awareness into two main modules: the context recognition module (SP1 FuturAAL-context) and the activity recognition module (SP2 FuturAAL-ego). That division responds to the idea that, to plan proper assistive strategies, the activities that a person is performing have to be understood, but also the context in which those activities are being performed. Hence, we need to coordinate two different subprojects with their own partial objectives in order to achieve a situational awareness module for future AAL systems which aim to work in non-structured environments.