Feature Extraction of EEG Signal Using Convolutional Neural Networks by Removing Artifacts

Abstract

Clinical depression is a neurological disease identifiable by the analysis of the electroencephalography signals (EEG). The electroencephalographic signals (EEG) are often polluted by many artifacts. Deep study models have been employed in recent years to denoise electroencephalography. The main difficulty in medical analysis is the extraction of true brain signals from the polluted EEG data. Noise reduction from recorded EEG data is very important for better brain disorder investigation. This paper proposed an effective EEG signal estimation model for the process of EEG signals. The proposed model uss the Morelette wavelet transformation model for the pre-processing of the EEG signal. With the pre-processed EEG signal model feature extraction is performed with the Convolutional Neural Network (CNN) for the EEG signal. With the pre-processed EEG signal model training and testing are estimated for the classification of the EEG signal. The EEG signal categorization was carried out utilizing characteristics derived from EEG data. Many characteristics have proven sufficiently distinctive for usage in all applications linked to the brain. The EEG may be categorized using a range of functions such as autoregression, energy spectrum density, energy entropy and linear complexity. However, various characteristics indicate varying strength of discrimination for different individuals or trials. Two characteristics are utilized in this study to enhance the performance of EEG signals. Techniques based on the neural network are used for the extraction of EEG signal. Classification methods include the Random Forest Classification. The model was tested using a random splitting method and 93.4 percent of the EEG signals were received accordingly.