next up previous contents
Next: Statistical tables Up: CASE STUDIES Previous: CASE STUDIES

Multichannel EEG signal

The electroencephalogram (EEG) [Lopes da Silva et al., 1986, Niedermeyer and Lopes da Silva, 1987, Nunez, 1981] consists of a set of signals measured with electrodes on the scalp. The pattern of changes in the signals reflects some large-scale brain activity; for example the occurrence of certain kinds of oscillation patterns is known to be correlated with certain vigilance states of the subject. In addition to brain activity, the EEG also reflects activation of the head musculature, eye movements, interference from nearby electric devices, and changing conductivity in the electrodes due to the movements of the subject or physiochemical reactions at the electrode sites. All of these activities that are not directly related to the current cognitive processing of the subject are collectively referred to as background activity below.

EEG measurements provide plenty of continuous-valued, time-dependent data, and an EEG-specific feature extraction procedure is needed for revealing interesting activity patterns. In Publication 1 the background EEG activity is visualized using the SOM, after extracting frequency-based features from the short-time spectra of the EEG signals of all channels. The 140-dimensional data items correspond to the short-time frequency content of the EEG in multiple bands and in multiple locations on the scalp.

The resulting map distinguished between different kinds of background activity. The different activity types were predominantly projected onto different areas of the map. Each of the areas was more or less connected, even though samples from many subjects were used. By visualizing the trajectory of successive data items it is possible to monitor changes in the background activity, and the types of activity underlying different map areas can be inspected by visualizing the reference vectors.

Sami Kaski
Mon Mar 31 23:43:35 EET DST 1997