Emotional Stress Recognition System for Affective Computing Based on Bio-Signals

Abstract

In this paper, we propose a new approach to classify emotional stress in the two main areas of the valance-arousal space by using bio-signals. Since electroencephalogram (EEG) is widely used in biomedical research, it is used as the main signal. We designed an efficient acquisition protocol to acquire the EEG and psychophysiological. Two specific areas of the valence-arousal emotional stress space are defined, corresponding to negatively excited and calm-neutral states. Qualitative and quantitative evaluation of psychophysiological signals have been used to select suitable segments of EEG signal for improving efficiency and performance of emotional stress recognition system. After pre-processing the EEG signals, wavelet coefficients and chaotic invariants like fractal dimension, correlation dimension and wavelet entropy were used to extract the features of the signal. So, by using independent-sample T-Test and Linear Discriminate Analysis (LDA), effective features are selected. The results show that, the average classification accuracy were 80.1% and 84.9% for two categories of emotional stress states using the LDA and Support Vector Machine (SVM) classifiers respectively. We achieved an improvement in accuracy, in compared to our previous studies in the similar field. Therefore, this new fusion link between EEG and psychophysiological signals are more robust in comparison to the separate signals.

Introductory Chapter: Emotion and Attention Recognition Based on Biological Signals and Images

“…Up to now, researchers and neuroscientists have studied continuously to improve the performances of the emotion and atention recognition systems (e.g., [1][2][3][4][5][6][7][8][9][10]). In spite of all of these eforts, there is still an abundance of scope for the additional researches in emotion and atention recognition based on biological signals and images.…”

Classification of Brain Activity in Emotional States Using HOS Analysis

“…All subjects had normal or corrected vision; none of them had neurological disorders. Each participant was examined by a dichotic listening test to identify the dominant hemisphere [9,14]. We used a Flexcom Infiniti biofeedback device for data acquisition.…”

“…The EEG signals, sampled at 256 Hz, were recorded from five channels (FP1, FP2, T3, T4 and Pz) placed on each subject's scalp according to the international 10-20 system. Each recording lasted about 3 minutes [9,14]. In our research, the stimuli to elicit the target emotions (calm-neutral and negatively excited) were some of the pictures.…”

“…We recorded psycho-physiological signals concomitantly in order to firstly recognize the correlated emotional state and then label the correlated EEG signal. More details of the data acquisition protocol can be found in reference [9,14].…”

Classification of Emotional Stress Using Brain Activity

“…In order to choose the best emotional stress correlated EEG signals, we implemented a new emotionrelated signal recognition system, which has not been studied so far (Hosseini, 2009;Hosseini et al, 2010c). We recorded peripheral signals concomitantly in order to firstly recognize the correlated emotional stress state and then label the correlated EEG signals.…”