Detection of Eye State using Brain Signal Classification with EBPTA and KNN Algorithm

Electroencephalograms (EEG) signals generally vary rapidly with time. These signals also have effect on the biotic lives which have brain. As the signals are rapid, their categorization becomes difficult. Most of the works regarding EEG signal categorization are highly domain specific. It requires a lot of time for the sub processes which are processing of signals and feature extraction. As EEG signals are analog, rapid changing, they have negligible Signal to Noise Ratio (SNR) value and hence this makes them vulnerable to any small disturbances or discrepancies. The proposed work is hence non-domain specific and has a comparison about the accuracies when the objective is achieved using Evolutionary Back Propagation Training Algorithm (EBPTA) and k-nearest neighbor (K-NN) algorithms. Convolutional operation is used for effective changing of specific data that can reveal the implicit spatial dependence of the Electroencephalography signals distribution. A dataset of 14 features and information about eye status is taken, and is tested, validated after training accordingly with two algorithms that is EBPTA and K-NN algorithms differently. The trails done prove that our work outperforms few other algorithms and results in accuracies of 98.94% and 96.99% for EBPTA and K-NN algorithms respectively, on chosen dataset with considerable resilience and time which makes them suitable for various range of problems which may be encountered in future.