A study of EEG signals modeling for different scent intensity levels

Entropy is a measure of information carried by a signal. It has also been used as a feature for the modeling or classification of signals. In this paper, we investigate the use of entropy for modeling EEG (Electroencephalography) based scent intensity levels. The paper examines three variations of the entropy design (i.e. entropy ratio, entropy difference, and entropy mean) and two other parameters namely root-mean-square (RMS), and Kurtosis. In order to derive the feature vectors, EEG signals are collected from 14 healthy volunteer human subjects on two levels of scent intensities. The results show that, EEG signals with higher scent stimulation have lower feature parameter values for all the five features considered. The feature vectors are also observed to clutter together for the same intensity level. A mathematical model using the five features is also proposed to represent or differentiate the intensity levels. An example of 3D visualization using three of the features considered is given to illustrate the modeling concept. In comparison to previous works, our paper focuses on the use of a mathematical model, involving variations of entropy as the features, to represent and differentiate EEG signals generated in response to different scent intensity levels.