A Two-Stage Channel Selection Model for Classifying EEG Activities of Young Adults with Internet Addiction (2016)

link to study

Advances in Neural Networks – ISNN 2016

Volume 9719 of the series Lecture Notes in Computer Science pp 66-73

Date: 02 July 2016

  • Wenjie Li
  • , Ling Zou 
  • , Tiantong Zhou
  • , Changming Wang
  • , Jiongru Zhou


Full scalp electroencephalography (EEG) recording is generally used in brain computer interface (BCI) applications with multi-channel electrode cap. The data not only has comprehensive information about the application, but also has irrelevant information and noise which makes it difficult to reveal the patterns. This paper presents our preliminary research in selecting the optimal channels for the study of internet addiction with visual “Oddball” paradigm. A two-stage model was employed to select the most relevant channels about the task from the full set of 64 channels. First, channels were ranked according to power spectrum density (PSD) and Fisher ratio separately for each subject. Second, the occurrence rate of each channel among different subjects was computed. Channels whose occurrences was more than twice consisted the optimal combination. The optimal channels and other comparison combinations of channels (including the whole channels) were used to distinguish between the target and non-target stimuli with Fisher linear discriminant analysis method. Classification results showed that the channel selection method greatly reduced the abundant channels and guaranteed the classification accuracy, specificity and sensitivity. It can be concluded from the results that there is attention deficit on internet addicts.


Channel selection Electroencephalogram (EEG) Internet addiction Oddball Power spectrum density Fisher linear discriminant analysis