EEG pattern identification for motor imagery based on 1DCNN-GRU

Due to unique secrecy, vividness, and unpredictability, electroencephalogram signals are regarded an efficient method of identification for security reasons. However, the EEG-based person identification method study is still in its infancy. Decrypting EEG signals and implementing EEG-based person identification is tough. From an application standpoint, this paper proposes a method employing a one-dimensional convolutional neural network and a gated recurrent unit network cascaded model for the identification, which can extract robust and rich spatio-temporal features from EEG signals efficiently and quickly. The test uses the Physionet EEG motor imagery dataset, which is available to the public and is made up of EEG from 109 subjects, to determine which electrodes work best during a screening task triggered by EEG. The experimental findings show that the identification rate of regions can be up to 99.86% at 16 electrodes (CP and P regions), which is improved in comparison to 64 electrodes. The proposed approach has been shown to make the identification of EEG-based persons easier for practical applications.