Hybrid Deep Learning Based PDOA Model for Prediction of BCI Task Using EEG Signals

Classification of motor imagery using electroencephalography (EEG-MI) is an integral part of the brain-computer interface (BCI), which helps those with mobility impairments connect with the outside world. Unfortunately, the EEG signal is complicated, highly dynamic, and has a poor signal-to-noise ratio, making deciphering the information difficult. It is still challenging to design a whole system architecture that can reliably extract high-level characteristics from EEG data. This research presents a novel model for BCI classification tasks that make use of EEG data. From the raw EEG data, we can pull out useful characteristics like spectrograms and cepstral frequencies. Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) are combined in a hybrid model (CNN+LSTM), which is then fed the data. In this study, we use the Prairie dog optimization Algorithm (PDOA) to determine the best values for the LSTM weighting parameters. In order to properly categorize signals coming from various channel locations, the suggested network was put into action. The outcomes were associated to those of the state-of-the-art. Because of this, the suggested strategy enhanced the accuracy of BCI categorization.