Overview of the EEG Signal Processing Technique

The description and overview of the classifiers, the different emotion models and the window filters (whose main purpose is to mitigate the impact of spectral leakage; this leakage is because of the discontinuity that is present in the time domain) are provided.

Is your time best spent reading someone else’s essay? Get a 100% original essay FROM A CERTIFIED WRITER!

Order now

SVM classifier

The primary role of the SVM classifier is to categorize emotions. For it to produce effective and efficient results it is imperative to ensure that the classifier has the following properties; great generalization characteristics, it should not be affected by overuse and it should be resistant and invulnerable to the various phenomena that mostly arise in the process of data analysis and data organization in high dimensional places. The general principle under which this classifier functions is that it recognizes the best or most favorable subspace(s) of one dimension less than their ambient space(s). It is in these spaces where the anticipated categorization errors of test samples will be reduced. The best subspaces are the ones which increase the margins. The reason for increasing the margins is to intensify and strengthen its generalization abilities. There are several components and proponents in this classifier. One is the regularization parameter C whose primary function is hosting of data points which are either very much bigger or smaller than the next nearest data points. Another of its function is to permit miscalculations and oversights in the training set. (Picard, 2000) The K-nearest neighbor (KNN) classification system has two main advantages; first, it is not a complicated technique and secondly, it is a powerful reliable technique. This happens to be instance-based learning classifier system whose function is to grouping and categorization of unfamiliar and unrecognized cases springing from a number of real valued functions which measure the similarity between two objects. The numerical value of K is chosen so as to get the most optimum levels of accuracy when it comes to the classification process. In this research, the ranges of numerical values which will be used for K are between two and five.

Model of emotionEmotions are cognitive, rational and cerebral (inner) states and as such they have a close relationship to a vast and broad variety of perceptions, thoughts, discernments, apprehension and behaviors. There have been so many discussion and debates in coming up with the best definition of emotions and the correct measure for its discrimination. Many researchers have been conducted to establish if emotions are mental actions and processes or whether they are non-cognitive. Currently, there are two frameworks which are used in the representation of theoretical emotion. These two are; discrete emotion model together with bi-dimensional emotion model framework.

In discrete emotional model, varying emotions such as joy, disgust, anger, shock, fear and happiness are generally pinpointed and recognized as basic emotions.

When it comes to the second model, the representation and categorization of emotions is through the use of a scale which contains several dimensions and aspects spread over valence and arousal. The advantage of the bi-dimensional emotion framework is that it is easily understood, not complicated and can be applied in many situations. (Chethan, P, 2002) Bi-dimensional model contains four major emotional divisions and these are; low arousal/low valence (LALV), low arousal/high valence (LAHV), high arousal/low valence (HALV) and high arousal/high valence (HAHV). In this research paper, six main categories have been chosen and these include; happiness, surprise, anger, disgust, sadness and fear.

Temporal windowIn regards to the recognition and identification of emotions, temporal aspects are very important. However, these have not been accorded the attention that they deserve. In a perfect hypothetical affective interface, detection and identification of the emotions of a particular person should be conducted in the shortest time possible. This swiftness enables effective decision making process in regards to the expectations of the individual. The optimal and correct temporal window size will be determined by the type of emotion together with the physiological signal. On most occasions, the time span of emotions ranges between half a second and four seconds (Amiri, 2011).

It is vital to get the size of the temporal window right so as to avoid problems which arise in the process of classifying emotions. The wrong size will ultimately lead to errors in the categorization process. Varying emotions can be experienced incase very lengthy or very short time spans are evaluated. Unluckily, a fixed proper window size which has the ability of obtaining the best and optimal emotion recognition has not been developed yet. As such, in this project, the size of temporal windows selected was two and four. These were applied in production of the EEG signals.

MATERIAL AND METHODS

This section is going to describe the method through which the EEG signals were acquired and collected. Data was collected from three male subjects and this was in regard to the aforementioned six categories of emotions, i.e., happiness, fear, disgust, surprise, sadness and anger. The names of the subjects were Ali, James and Marty and the machine used in this process is referred to as a Flex comp machine. It is important to point out that the three subjects had no psychiatric, structural, biochemical or nervous system abnormalities. (Cox, M. (2002) for research ethical adherence, the three subjects provided their consent before the commencement of the recording process. Before starting the experiment, the respondents were informed about the design, its implications and its goals and objectives. The experiment took place during the subjects free leisure time.

The Nevus Electroencephalogram was used in the identifi...