Physiological wireless sensor network for the detection of human moods to enhance human-robot interaction

Although it is already possible to issue utility services that use robots, these are still not perceived by society as capable of actually delivering them. One of the main motivations is the lack of a human-like behaviour in the interaction with the user. This is displayed both at physical and cognitive level. This work investigates the optimal sensor configuration in the recognition of three different moods, as it surely represents a crucial element in the enhancement of the human-robot interaction. Mainly focusing towards a future application in the field of assistive robotics, electrocardiogram, electrodermal activity and electroencephalographic signal were used as main informative channels, acquired through a wireless wearable sensor network. An experimental methodology was built to induce three different emotional states by means of social interaction. Collected data were classified with six supervised machine learning approaches, namely decision tree, induction rules and lazy, probabilistic and function-based classifiers. The results of this work revealed that the optimal configuration of sensors which maximizes the trade-off between accuracy and obtrusiveness is the one surveying cardiac and skin activities. This sensor configuration reached an accuracy of 87.07% in the best case.