A Novel Deep Learning Framework for High-Gamma EEG Signal Classification Using Hybrid Compact-CNN-LSTM Network

Electroencephalography (EEG)-based motor imagery (MI) classification plays a crucial role in developing efficient brain-computer interface (BCI) systems. However, decoding EEG signals is challenging due to their low signal-to-noise ratio and complex temporal dynamics. In this study, a novel compact hybrid deep learning framework combining one-dimensional convolutional neural networks (1D-CNN) with a long short-term memory (LSTM) network (CCLNet) is proposed to classify MI-related EEG signals. Raw EEG data from two MI classes were segmented into non-overlapping 4-second trials and recorded across 50 channels. The model first employs 1D-CNN layers to extract spatial features from the EEG signals, followed by an LSTM layer to capture temporal dependencies. The classification layer outputs two classes corresponding to left-hand and right-hand MI tasks. Experimental evaluation demonstrates that the proposed CCLNet model achieves robust performance, with an average accuracy of 98.25%, outperforming traditional machine learning baselines. These results highlight the effectiveness of combining temporal and spatial feature learning for EEG classification and suggest potential for deployment in real-time BCI systems.