ABSTRACT
Emotion is an important part of people's daily life, particularly relevant to the mental health of people. Emotional diagnosis is closely related to the nervous system, which can well reflect people's mental conditions in response to the surrounding environment or the development of various neurodegenerative diseases. Emotion recognition can help the medical diagnosis of mental health. In recent years, emotion recognition based on EEG has attracted the attention of many researchers accompanying with the continuous development of artificial intelligence and brain computer interface technology. In this paper, we carried out a comparison on the performance of three deep learning techniques on EEG classification, including DNN, CNN and CNN-LSTM. DEAP data set was used in our experiments. EEG signals were transformed from time domain to frequency domain first, and then features are extracted to classify emotions. From our research, it shows these deep learning techniques can achieve good accuracy on emotional diagnosis.
