Abstract
Understanding human behaviour in an automatic but also non-intrusive manner, constitutes an important and emerging area for various fields. This requires collaboration of information technology with humanitarian sciences in order to transfer existing knowl- edge of human behaviour into self-acting tools to eliminate the human error. This work strives to shed some light in the area of Mobile Social Signal Processing by trying to understand if today’s mobile devices, given their advanced sensing and computational capabilities, are able to extract various aspects of human behaviour. Although one of the core aspects of human behaviour are social interactions, current tools do not pro- vide an accurate, reliable and real-time solution for social interaction detection, which constitutes a significant barrier in automatic human behaviour understanding. Towards filling the aforementioned gap in order to enable human behaviour under- standing through mobile devices, particular contributions were made. Firstly, an interpersonal distance estimation technique is developed based upon a non-intrusive opportunistic mechanism that solely relies on sensors and communication capabilities of off-the-shelf smartphones. Secondly, based on user’s interpersonal distance and relative orientation, a pervasive and opportunistic approach based on off-the-shelf smartphones for social interaction detection system is presented. Leveraging information provided by psychology, analytical and error models are proposed to estimate the probability of people having social interactions. Then, to showcase the ability of mobile devices to infer human behaviour, a trust relationship quantification mechanism is developed based on users’ behavioural traits and psychological models. Finally, a prediction and compensation mechanism for the device displacement error that leverages human loco- motion patterns to refine the device orientation is introduced. The above contributions were evaluated through experimentation and hard data collected from real-world environments to prove their accuracy and reliability as well as showing the applicability of the proposed approaches in daily situations. This work showed that mobile devices are able to accurately detect social interactions and further social and trust relationships among people, despite the noise induced in real-world situations. Close collaboration between informatics and social sciences is imperative, to overcome the significant barrier in the development of human behaviour understanding. This work could constitute a fundamental building block, as the computational power and battery autonomy of mobile devices increases, for the development of novel techniques towards understanding human behaviour, by including multiple behavioural traits and enabling the creation of socially-aware information systems.