An enhanced deep learning model based on smoothed pseudo Wigner-Ville distribution technique for emotion recognition with channel selection

Abstract

Emotion recognition is indispensable in cognitive action processes throughout human life. In the previous studies, the emotion recognition problem is approached with linear signal processing methods. Therefore, this article proposes non-linear feature extraction method using smoothed pseudo Wigner-Ville distribution (SPWVD). An eclectic deep learning model 1D-CNN-GRU-BiLSTM is developed by combining gated recurrent unit (GRU) and bidirectional long short-term memory (BiLSTM) at the end of the one-dimensional convolutional neural networks (1D-CNN). The experimental results have shown that the proposed model obtained the highest success rates of 91.22% for the SEED dataset and 90.76% and 91.22% for the DEAP dataset during the valence and arousal axes, respectively. Moreover, in channel selection, the proposed model process is reached accuracy rate 84.76% for SEED dataset and 89.07% (valence-axis), 89.41% (arousal-axis) DEAP dataset for 16-EEG channels. The 1D-CNN-GRU-BiLSTM architecture demonstrates superior performance in EEG-based emotion recognition, outperforming existing methods by yielding higher recognition rates.

Keywords

• Emotion recognition
• Wigner-Ville distribution
• 1D-CNN
• GRU
• BiLSTM
• Channel selection