Computer Science Department
School of Computer Science, Carnegie Mellon University
Using a Low-cost EEG Sensor
Bao Hong Tan
The ability to detect mental states, whether cognitive or affective, would be useful in intelligent tutoring and many other domains. Newly available, inexpensive, single-channel, dry-electrode devices make electroencephalography (EEG) feasible to use outside the lab, for example in schools. Mostow et al.  used such a device to record the EEG of adults and children reading various types of words and easy and hard sentences; the purpose of this experimental manipulation was to induce distinct mental states. They trained classifiers to predict from the reader‘s EEG signal the type of the text read. The classifiers achieved better than chance accuracy despite the simplicity of the machine learning techniques employed.
Their work serves as a pilot study for this thesis and provides the data set for all analyses in this work. This thesis further explores the properties and temporal structure of the EEG signal with the aim of improving the accuracy of detecting mental states, with a focus on "easy" and "hard", that is, when the user is having difficulty with reading the given text or not. The EEG signals associated with the word stimuli are analyzed for the existence of event-related potentials (ERP) that could distinguish the word type, which in turn could be exploited in classification. The EEG signals for the sentence stimuli are subjected to various feature extraction methods and temporal manipulations. This thesis demonstrates the potential of exploiting the temporal structure of EEG signals in detecting mental states.