Challenges in data collection in real-world environments for activity recognition

Detecting and recognizing activities of daily living is an important part of ambient assisted living (AAL) systems. This part of the system has the highest impact on the overall system efficiency because it directly provides insights into the user's health state. One of the main challenges that AAL systems are facing are the privacy concerns and the intrusiveness of the sensors that are being deployed. In an ideal scenario, an aged person should be able to continue his or her normal life without noticing that they are being monitored. Another issue for such systems is the data collection. The current approaches usually use data generated in labs and data from end-users users is usually unavailable due to ethical concerns and the inability to deploy them in their living environments. Publications that rely on real-life scenario data are scarce. In this paper, we present the challenges one faces when trying to produce a sound dataset for further analysis and suggest ideas for overcoming them.