Emotional Recognition of Bi-direction GRU Models Based on EEG-Based Biometric Signals

초록

Until now, several studies have been conducted for human-robot interaction, and the emotion recognition part is a fundamental technology for human-robot interaction. In recognizing emotions, it is largely divided into a discrete method and a continuous method by using bio-signals. However, since the discrete method recognizes and expresses only simple emotions, it is an emotion recognition method that is unfavorable for real-time emotion recognition. Therefore, in this paper, the bio-signal of MAHNOB-HCI, the published autonomic nervous system signal data, uses the continuous method of real-time emotion recognition and the Bi-direction Gated Recurrent Units(GRU), a modified model of Long Short-Term Memory(LSTM) used to recognize data continuously. We propose a Valence-based recognition result by applying Electroencephalogram(EEG) signal data.