Using Squeeze and Excitation Attention Mechanism to enhance the classification accuracy of EEG Motor Imagery Signals

The communication process between the human brain and the computer is an important thing these days for various reasons, as humans can now issue commands to electronic devices by reading the electrical signals of brain and translating them into programming commands that the computer understands and interprets. The Brain-Computer Interface (BCI) technology or system is considered one of the best technologies in this field, where brain signals or Electroencephalogram (EEG) motor imagery (MI)  are taken as input to this system and translated into commands on the computer, scientific research in the field of artificial intelligence and machine learning, especially the so-called brain-computer interfaces, is no longer directed only to people with motor disabilities, but rather has become directed for general use in order to improve Experience our use of computers and human communication with computers better. in addition, to assist disabled people, control smart devices or environments, and even augment human capabilities. In this article, we propose a squeeze and excitation attention-based convolutional neural network, this model SE-CNN utilizes multiple techniques to boost the performance of MI classification with a relatively small number of parameters, SE attention block enhances the generalization and robustness of the network. In the BCI Competition IV-2a dataset, our proposed model has obtained good experimental results, achieving accuracies of 83.44% for subject-specific performance on the BCI-2a dataset using the same original competition division (hold-out approach: 50% training trials and 50% test trials). In addition, extensive ablative experiments and fine-tuning experiments were conducted.