Selected channel based multiclass emotion classification from wearable human brain EEG signal

Emotion recognition is a crucial issue in human-computer interaction, and EEG (electroencephalography) plays a significant role in deciphering human emotions based on physiological data. However, the complex and non-stationary nature of EEG signals, coupled with redundant information from multi-channel recordings, poses challenges in accurate emotion classification. To address this, we propose a hybrid 1DCNN-Bi-LSTM model that integrates spatial feature extraction (1DCNN) with temporal dependency learning (Bi-LSTM), enhancing the robustness of emotion classification. Furthermore, we present a channel selection mechanism to find the most pertinent EEG channels for emotion recognition, hence lowering computing complexity without compromising accuracy. With the chosen-channel model (8 channels) attaining 85.16 % accuracy, a notable improvement over standard full-channel approaches, experimental results on the DEAP dataset show that the suggested methodology provides significant performance gains. This work fits wearable devices and real-time affective computing systems since it offers a scalable and effective method for EEG-based emotion recognition.