EEG data preprocessing, prediction, feature extraction/selection and classification - Dissertation Example

This is indeed a step forward from the traditional use of EEG, where its function was limited to the assessment of a person’s brain activity. Such advances have become a reality with the identification of new neural signals, voluntary EEG patterns and creation of the brain computer interface or the BCI (Barreto et al, 1996, pp 73). The automated neural network system based spike detection systems can work utilizing either of the two components (Kalayci and Ozdamar, 1995, pp 160 and 162). It can either utilize the extracted EEG parameters or it can use the raw EEG signals.

The proper selections of the data, in this case the spikes and slopes, lead to more accuracy in the final results (Kalayci and Ozdamar, 1995, pp 160 and 162). There are many reasons why the use of new methods is advocated. For example, feature selection has been a very effective tool in communicating with patients who are otherwise paralyzed, and cannot carry out physical motion to show responses (Deriche and Al-Ani, 2002, pp 2961). The feature selection helps in creating a patient code to understand the reaction that he or she is providing.

Utilizing these signals from the brain can help create a brain computer interface or BCI, which allows for correct interpretation (Deriche and Al-Ani, 2002, pp 2961). . This wave helps in channeling the neural processes of one particular movement, which leads to the creation of a series of reactions and neural synaptic activity at the related junctions. The phenomena have been termed as “Event Related Desynchronization” or ERD (Barreto et al, 1996, pp 73). In this regard, the classification of these activities can be carried out through classification of the readiness potentials, using a 12 electrode subset of 10-20 International Electrode System.

Readiness potentials are the “transient activation of the current dipoles with characteristic location and temporal pattern of activation” (Barreto et al, 1996, pp 74). For better analysis of the EEG, a technology known as the Artificial neural networks was developed (Guo et al, 2009, pp 178). This technique has been shown to improve the assessment of the brain signals better, by identifying complex patterns and classifying different waves with ease (Guo et al, 2009, pp 178). The system uses a variety of algorithms, the most notable of which is the back propagation algorithm.

This system ensures the removal of errors from the data and helps in creating a clear idea of the changes taking place in the EEG (Guo et al, 2009, pp 178). Feature extraction stage is among the initial stages of automated analysis. Feature selection can be based on the concept of “mutual information”, which allows measurement of the strength between different variables (Deriche and Al-Ani, 2002, pp 2962). These can be of many types, examples of which include spectral analysis based on parametric (AR modeling), non parametric (FFT) and multi-scale (WVT).

All of these models are based on the calculations of the various powers of