Deep Learning for BCI Signal Decoding
Harpreet Kaur Channi, Surinder Pal Kaur, Ramandeep Singh Sandhu · 2025
BCIs enable direct brain-to-device communication by decoding brain signals, with applications in assistive technologies, neurorehabilitation, and human-computer interaction. This study explores deep learning models, including CNNs, RNNs, and hybrid architectures, for BCI signal decoding. A synthetic EEG dataset was used to benchmark these models against traditional classifiers like SVM and LDA, focusing on accuracy, latency, and generalization. Preprocessing techniques such as noise reduction, signal normalization, and feature extraction were applied to enhance signal quality. Results showed CNNs and hybrid CNN-RNN models outperformed traditional classifiers with greater accuracy and robustness. Deep learning effectively addressed challenges like low signal-to-noise ratio, subject variability, and class imbalance. The study also demonstrated practical applications in assistive device control and cognitive monitoring, emphasizing deep learning's potential to advance BCI technology while underscoring the need for validation on larger datasets.