Hybrid Neural Network Genetic Algorithm Framework for EEG and ECG Signal Classification

Clinically relevant information is extracted from biomedical signals such as electroencephalogram (EEG) and electrocardiogram (ECG) by digital signal processing (DSP) techniques, which are pivotal. A hybrid framework that synergizes Convolutional Neural Networks (CNNs) with Genetic Algorithms (GAs) is introduced in this paper to enhance the classification accuracy of EEG seizures and ECG arrhythmias. Wavelet transformers are employed for noise-robust feature extraction, and a GA is utilized for automated hyperparameter optimization of a CNN-BiLSTM architecture. The proposed model is evaluated on the CHB-MIT EEG dataset (23 subjects) and the MIT-BIH ECG dataset (47 subjects), achieving 96.3% and 95.8% accuracy, respectively, and outperforming ResNet-18 and SVM baselines. The significance of improvements is confirmed by statistical tests (Wilcoxon signed-rank, p<0.01). Inference latency is reduced by 38% by the optimized model, making it suitable for edge deployment in real-time diagnostic systems.