Design and Development of Cognitive Learning Framework for Visual Image Reconstruction Using Brain Evoked Signals

Decoding perceptual experiences from brain activity has emerged as a critical area of research in neural computation, driven by its potential applications in advanced technologies and cognitive interfaces. Electroencephalography (EEG), a non-invasive technique with high temporal resolution, is widely utilized for capturing brain-evoked activity. However, existing approaches face challenges in effectively modelling the nonlinear and dynamic characteristics of neural signals, thereby limiting recognition performance under complex visual conditions. To address these limitations, this work proposes a hybrid deep learning framework that integrates EEGNet - a convolutional neural network (CNN) tailored for EEG signal analysis with a recurrent neural network (RNN) to model temporal dependencies. The model is trained and evaluated using the Mendeley Natural Image EEG dataset, which contains brain activity recordings elicited by natural visual stimuli. EEGNet is employed for extracting discriminative spatial features, while the RNN captures temporal sequences associated with visual perception. The proposed framework achieves a classification accuracy of 99.4 %, with a precision of 99.1 %, recall of 99.3 %, specificity of 99.2 %, and F1-score of 99.2 %, demonstrating its robustness and generalizability. The experimental results confirm that the hybrid EEGNet-RNN model surpasses existing methods in spatial-temporal EEG decoding, validating its efficacy for cognitive computing and brain computer interface applications.