Research Of Correlations Between EEG,Eye Movement Data And Attention,Depression

In recent years, there has been an increased interest in studying brain activity during real-world experiences. People’s moods heavily influence their ways of communicating and acting, as well as productivity, and also play a significant role in learning process. However, detecting students’emotion and attitude without interfering is usually a complex task, so practices are often difficult to carry out, especially in distance education program.Our work mainly focused on two kinds of learning related emotions, attention and depression. Most existing studies obtain affective information through speech, motion, gesture, facial expression, etc. Designing affective learning experiments on learners while introducing limited disturbances on learners is one of the challenges. New techniques need to be introduced to rich the ways of deeper understanding learners’ affect such as EEG and eye movement data approaches.As the biological information recorded during human activity, EEG (Electroencephalography) and eye-movement data were proved to have close relationships with affect during learning process. Based on the previous work, this thesis conducts a systematic analysis of relation between EEG/eye-movement data and attention/depression. We hope our work could afford us a new way of understanding students’ affect so as to trigger innovative thinking and enhance creativity in learning process.This thesis applies algorithms of data mining and signal processing to EEG data and eye movement data, with the aim of recognizing two kinds of affect that has close relation with learning process, attention and mild/moderate depression. The main contributions of this thesis are as follows:(1) This thesis investigates the influence of individual difference on EEG data. Results show that it is hard to find obvious rule relating to the appearance of EEG features. This thesis proposes a combination of CFS+kNN for EEG data processing in attention recognition. It was found that CFS+kNN had a much better performance, with the shortest running time and the highest correct classification rate(CCR), giving the CCRs of 83.11% and 88.44% for the valence dimension divided into 5 classes and 3 classes respectively. The results also indicated that the use of CFS can highly reduce running time and enhance the CCRs in EEG signal processing, which makes it suitable for implementation of EEG-based real-time system.(2)From the source imaging location method using EEG signals, it was found that Within-group comparison of EEG data showed higher theta activity in BA6 (Brodmann area) and higher alpha activity in BA38 also revealed the mild/moderate depression paid more attention to the negative face expressions, and the temporal pole suggested to be the first dysregulated place appeared in mild/moderate depression.The results of Brain Electrical Activity Mapping indicate that there is great difference in beta and low gamma bands between subjects with mild depression and normal controls. And coherence analysis show that in gamma band the global EEG coherence of subjects with mild depression was significantly higher than that of normal controls.(3) Topographic of band power show that, between the frenquency of 25 and 30Hz, beta wave of mild depressed subjects comes out to be more stronger than normal controls, and low gamma in frontal region shows the same pattern in emotional picture and silent picture block. Coherence matrix and topographic of clustering coefficient results show that in low gamma band, the group effect was mainly found in the frontal region of brain, this was also supported by the statistic analysis.