Enhancing Human Emotion Recognition with Long Short-Term Memory (LSTM) and Adaptive Adam Optimization (AOA) of EEG Signals

Abstract

Electroencephalogram (EEG) data-based emotion recognition has acquired popularity in recent years. Because EEG signals are noisy, non-linear, and non-stationary, it is challenging to develop an intelligent framework capable of delivering high accuracy for emotion recognition. Human emotion recognition is a critical component of human-computer interaction and assessing emotional well-being. In this paper, we describe a new method for increasing the accuracy of human emotion recognition using EEG data. Adaptive Adam optimization is used to enhance the performance of Long Short-Term Memory (LSTM) neural networks by utilizing their power. Readings of the EEG, which record brain activity, provide crucial insight into emotional states. EEG data are utilized to extract intricate patterns and temporal connections using LSTM networks, which are well-known for their ability to efficiently characterize sequential data. To further optimize training and convergence, we provide the adaptive Adam Optimization Algorithm (AOA), which dynamically modifies learning rates during training to improve the model's ability to capture minute variations in EEG patterns. Experiment results from comprehensive evaluations on benchmark datasets indicate that our method is effective at improving the accuracy of identifying human emotions from EEG signals. The proposed model, LSTM-AOA, outperforms conventional methods and demonstrates exceptional proficiency in dealing with the complex temporal dependencies inherent to emotional states. © 2023 IEEE.