EEG-Based Emotion Recognition In Chinese Emotional Words

Affective computing is one of the key technologies to achieve high-level human-computer interaction, and it is becoming a growing concern in the field of artificial intelligence. The research contents include emotional mechanism, emotional information acquisition, emotional pattern recognition, emotional synthesis and expression, emotion transfer and communication, etc. Among them, the emotional recognition is a key problem of emotional computing research. It mainly includes speech, facial expression, text, posture and physiological signal recognition, etc. The research on physiological signal to affect recognition study is the most difficult. The main research object of physiological signal based emotional recognition includes electroencephalograph(EEG), muscle power, Electrocardiograph(ECG), skin, skin temperature, conductive photoelectric pulse, respiration signals, etc.EEG is the hundreds of millions of neurons within the brain activity in a comprehensive reflection of the cerebral cortex, and it can directly reflect the brain’s activities. Different state of mind and mood changes would reflect different EEG in a variety of different locations in the cortex. Therefore, the EEG contains a wealth of useful information. How to effectively deal with these signals and extract useful information for the research of identifying the emotional state is important. However, since the EEG is kind of non-stationary random signal with very complex mechanism of generation, and the data collection process is very complex, and is easily susceptible to external environment and other physiological signals such as ECG, EMG and so on, therefore, the research on emotion recognition is very scarce whether at home or abroad compared with other physiological signals.The main contents of this paper are as follows:(1) Selection of Chinese emotional words:First, we artificially select 3750 words from the’Modern Chinese Dictionary’, and based on emotional corpus of our laboratory,945 commonly used words with higher word frequency are selected. Finally,20 words with highest emotional intensity are chosen by questionnaires.(2) Data collection:we collect EEG signals from seven subjects using ESA Pro/Basic system. All subjects are graduate and doctoral students in Tokushima University in Japan.(3) Feature extraction:we use time domain analysis and frequency domain analysis to extract a combination of EEG features. They are cross-correlation, principal component analysis and a waves share.(4) Classification and evaluation:we use the support vector machine (SVM) and linear discriminant analysis (LDA) as a model of emotion recognition classifiers to verify the feasibility of the experimental hypothesis. For both classifiers, the best accuracy rate reached 51.47% and 65%.The main purpose of this study is to investigate the relation between emotional responses and EEG data gathered by giving text information as stimuli. We first select 20 Chinese emotional words with high emotional intensity and high frequency for our experiment. And then we chose cross-correlation coefficient, principal component analysis and ratio ofα-wave for feature extraction, and finally we use support vector machines and linear discriminant analysis classifier for our emotion classification task. From the experimental results, we can see that it is effective to extract those three EEG features for emotion recognition, and it is feasible to obtain EEG signals through using association-based methods for emotion recognition.