Eeg-based emotion recognition with brain network using independent components analysis and granger causality

With the continuous development of brain imaging technology, it has become a hot area of neuroscience and information technology to research the human emotion changes, cognitive status and psychiatric disorders. In recent years, any smart device can be used as a terminal sensor in the Internet of Things for information interaction. It will be the new research aspect for Brain-Computer Interface(BCI) to regard the human brain (the most intelligent “device”) as a terminal sensor in the Internet of Things and to construct the network based on the human brains (we name it as Internet of Brains). In this paper, a model of wearable affective computing was proposed for discriminating different emotional states and constructing the Internet of Brains, by means of effective connectivity of EEG-based brain network. Firstly, we proposed a rational emotion-induced psychological experiment to collect the EEG data under different emotional states. Then, Independent Component Analysis (ICA) was used to decompose different independent components based on different emotional states; Granger Causality Analysis (GCA) was utilized to detect the interactive dependencies between each independent component in order to construct the causal connectivity brain network (CCBN); Dynamic characteristics, including causal density and causal flow of the CCBN, were extracted based on Graph Theory. Finally, the corresponding law between characteristics of EEG pattern and “inner” emotional state was discovered to establish affective computing model. Furthermore, the model of wearable affective computing was constructed based on above law with the portable EEG acquisition device, and prototype system of wearable affective computing based on Internet of Brains was achieved for BCI.