Effective Cognitive State Monitoring Using FBNN Architecture

The integrated Electroencephalogram (EEG) signal processing and deep learning for robust cognitive state classification is the idea behind this paper. The EEG data that are received are first preprocessed. To preserve the accuracy of the data, noise artifacts from the raw EEG signals are removed during the preprocessing stage. The decomposed signals are subjected to statistically feature extraction in order to extract pertinent properties and produce an accessible summary of the data. By exposing the frequency domain data, spectral estimating improves the analysis even more and makes it possible to identify certain patterns linked to different cognitive states. Using deep learning methods for cognitive state classification is the last stage of the recommended methodology. Convolutional Neural Network (CNNs) are used because they are able to identify spatial connections in the data, while, Feed Forward Back Propagation Neural Networks (FFBNNs) have an architecture that is flexible enough to understand intricate temporal interactions. Mat lab software is utilized in the implementation of this paper. The comparison diagram between the algorithms as find for accuracy, sensitivity and specificity. The accuracy level of the obtained CNN and FFBNN at signal 120 is 91.7% and 92.8% individually.