A New Deep Learning Architecture Based on LSTM and Wavelet Transform for Epileptic EEG Signal Classification

Epilepsy poses a significant risk to human health. Numerous machine learning (ML) have emerged to address this health challenge. Deep learning (DL), particularly recurrent neural networks (RNNs), has proven highly effective in distinguishing EEG patterns from epileptic brain activity. This paper introduces an innovative architecture DL approach, combining wavelet transform and long short-term memory (LSTM) networks, to improve accuracy and efficiency in epilepsy classification. Many experimental results have been applied using various epileptic datasets and diverse metrics such as accuracy, precision, and recall among others to assess the effectiveness of our proposed solutions. In almost all of the cases, our approach demonstrates its performance in improving seizure detection accuracy and contributing to the field of epileptic seizure recognition.