Relevance of EEG input signals in the augmented human reader

Abstract

This paper studies the discrimination of electroencephalographic (EEG) signals based in their capacity to identify silent attentive visual reading activities versus non reading states.The use of physiological signals is growing in the design of interactive systems due to their relevance in the improvement of the coupling between user states and application behavior.Reading is pervasive in visual user interfaces. In previous work, we integrated EEG signals in prototypical applications, designed to analyze reading tasks. This work searches for signals that are most relevant for reading detection procedures. More specifically, this study determines which features, input signals, and frequency bands are more significant for discrimination between reading and non-reading classes. This optimization is critical for an efficient and real time implementation of EEG processing software components, a basic requirement for the future applications.We use probabilistic similarity metrics, independent of the classification algorithm. All analyses are performed after determining the power spectrum density of delta, theta, alpha, beta and gamma rhythms. The results about the relevance of the input signals are validated with functional neurosciences knowledge.The experiences have been performed in a conventional HCI lab, with non clinical EEG equipment and setup. This is an explicit and voluntary condition. We anticipate that future mobile and wireless EEG capture devices will allow this work to be generalized to common applications.

User experience evaluation through the brain's electrical activity

“…Another study [9] used EEG to determine change in brain activity during reading and non-reading states. A series of presentation slides, with some having text and some blank, were presented to participants for 30 seconds in the case of text slides and 20 seconds in the case of blank slides.…”

Practical Neurophysiological Analysis of Readability as a Usability Dimension

“…The selected and used computational techniques are also briefly enumerated, as a more detailed explanation and discussion of the processing techniques can be found elsewhere [20,21].…”

“…The feature selection can also be based on relevance analysis. This analysis ranks features according to their contribution to the discrimination of the classes under observation [21].…”

“…PSD measures the energy of the signal in a certain frequency [21]. It is a frequency feature, so to get its variation over time, the PSD is determined in one second length rectangular windows, overlapped in 0.5 seconds.…”

“…The feature selection can also be based on relevance analysis. This analysis ranks features according to their contribution to the discrimination of the classes under observation [21].All the results presented and discussed in this paper were obtained using PSD (Power Spectrum Density) feature extraction and an SVM (Support Vector Machines) classifier. PSD measures the energy of the signal in a certain frequency [21].…”