A physiological signal processing system for optimal engagement and attention detection

This paper proposes a computer aided system that aims to measure and interpret physiological signals so as to assess the attention/engagement level of a person during cognitive based activities. In this study, ECG (Electrocardiogram), HF (Heat Flux) and EEG (Electroencephalogram) signals were collected from 8 subjects. The subjects were made to watch a series of videos which demanded contrasting engagement levels. On the collected ECG data, Discrete Wavelet Transform (DWT) is applied to the raw signal and multiple features are extracted. Features from HF were also obtained. In EEG signals, different band components were first extracted upon which DWT is applied to extract numerous features. Finally machine learning techniques were employed to classify the extracted features into two categories of `attention' and `non-attention'. The results show success in distinguishing `attention' vs. `non-attention' cases by processing acquired physiological signals.