Sensors and Wearable-based Activity Recognition and Behaviour Analysis for Users' Quality of Life Improvement

The focus of the thesis is on sensors and wearable-based activity recognition and behaviour analysis for Users' quality of life improvement. The topic has been addressed considering two different fields of study, one concerning \ac{BECM} Systems while the other is more concerning healthcare and well-being aspects. Indeed, it has been largely demonstrated how human behaviour could impact, on different levels, the quality of life. Some behaviours directly affect individuals' life, not only considering their psycho-physical health state but also as a result of the big impact that human behaviour has even on Earth's life. Therefore, the main idea of this thesis is to analyse and monitor how human habits can have an impact on people's quality of life, inside their homes and also during their everyday activities outside the home. With this scope in mind, this thesis presents three different systems and prototypes that, thanks to different kind of sensors and devices, can understand, recognise, and learn users' habits and act accordingly to their preferences. People often undertake incorrect behaviour without realising it, so the final purpose of these systems is to provide information and advice to users in order to make them aware and correct some of these wrong behaviours or encourage positive ones. The first system is based on a sensor network inside a Smart Home that recognises and predicts all the activities that are going on inside the home and, considering energy prices and user preferences, can schedule the appliances of the home in order to guarantee energy savings while at the same time ensure the comfort of the inhabitants. The second system is an appliance power profiling system that analyses the power consumption data collected by smart meters, identifies which features are most relevant for the specific appliance and extracts the set of power consumption profiles that are associated with each appliance. This kind of solution is necessary so that the Smart Home system can learn and know the power consumption profiles over time of the appliances present in the building. Finally, the third system is concentrates more on understanding how some activities, especially those related to fit activities and sleeping habits, may be reflected in how the user evaluate their own well-being and stress levels. This system makes use of simple data collected thanks to commercial wristbands and the study is focus on find a correlation between this activity data and the score obtained from a self-evaluation questionnaire about the well-being as users themselves perceive it. The results obtained and illustrated in this thesis show how a more conscious use of energy inside buildings, achieved thanks to accurate scheduling of household appliances, can guarantee energy savings while considering the user's comfort and preferences. The study on appliance profiling made it possible to recognise different usage profiles relating to specific appliances in order to have different consumption profiles, each represented by a reference consumption profile that can be later used for making consumption predictions or for giving advice to users. The system based on common and commercial smart objects of daily use shows that a strong correlation can be found between simple data obtained from these devices about daily activities and the level of stress and well-being of users.