Compressing electroencephalography (EEG) signals is crucial for reducing data storage and transmission costs in real time healthcare monitoring and brain computer interface systems. This study proposes an improved EEG compression framework based on a convolutional auto encoder (CNN-AE) with integrated preprocessing. Unlike traditional approaches, this proposed framework systematically investigates the impact of preprocessing strategies on the performance of deep learning-based compression models. Specifically, we introduce a Gaussian filtering stage to enhance signal quality before encoding. The suggested  framework  was evaluated using the CHB-MIT EEG database ,the prepressing stage involves normalization, segmentation, and then extraction of features, followed by the application of a CNN based auto encoder for processing dimensionality reduction on EEG signals which is constructed in layers. Various learning rates, segment lengths, and training epochs were conducted for the process' overall framework robustness evaluation of the model. Mean Squared Error (MSE), Percentage Root Mean Square Difference (PRD), and Peak Signal-to-Noise Ratio (PSNR) are used for the model's evaluation on reconstruction performance. Experimental results demonstrate the superiority of the proposed framework (combining a Gaussian filter and a CNN-AE) over the baseline model (CNN-AE with band-pass filtering). With a compression rate of 8:1, the suggested method has achieved an MSE less than 8.1 × 10^-6, a minimized relative measure of Distortion (PRD) below 1%, and superior reconstruction quality (PSNR over 50 dB). This also proves to the evaluation that the Gaussian smoothing method is superior and that the compression is superior in overall and representation stability. The proposed framework offers an effective solution for signal compression with high computational efficiency, and its scope can be expanded to include practical applications in the field of biomedical signal processing.

EEG Compression, CNN Auto encoder, Gaussian Filter, Reconstruction Quality, Deep Learning for Biomedical Signals,CHB-MIT EEG database