Feature Optimization of Motor Imagery EEG Classification using Evolutionary Algorithms and Machine Learning for ALS Disease

Abstract

The accuracy and precision of ALS disease detection depend on EEG classification. The high dimension of EEG data and complex structure of features degraded the performance of ALS disease detection. The feature extraction cum optimization process is better ways to achieve high accuracy and precision. The motor imagery is frequency-time series data in the mode of the signal. The brain-computer interface (BCI) uses electroencephalography or other modes for measuring brain activity. The recorded and collected signals analysed and find the abnormal condition of acute diseases such as ALS, brain stroke, and many more. Various frequency-based transform functions are used for the extraction of features, such as FFT, DCT, DWT, and many more transform functions. The CSP (common spatial pattern) is the most dominated feature extraction method for EEG feature extraction. The diverse nature of features needs features optimization process for a better process of classification. Swarm intelligence and evolutionary algorithms play an essential role in the process of feature optimization newline newlineThe selection of feature component in motor imagery EEG is a big issue due to the complex structure of features. the performance of classification depends on the selection of feature component. The selection of features components reduces the dimension of features space of motor imagery EEG. This dissertation proposed hybrid feature selection components based on firefly and ant colony optimization. The firefly algorithms represent the time-frequency component as a feature attribute of non-stationary signals. The ant colony optimization identifies the distinct behaviors of features component for the process of classification. The derived support vector machine is multi-level Kernel based support vector machine for the process of classification of EEG data. newline newlineAnother proposed ensemble-based classifier for the prediction of critical disease. The extraction of features of EEG signals is also challenging task, for the extraction of features used wav