Abstract
Several lobes in the human brain are involved differently in the arousal, processing and manifestation of emotion in facial expression, vocal intonation and gestural patterns. Sometimes people suppress their bodily manifestations to pretend their emotions. Detection of emotion and pretension is an open problem in emotion research. The chapter presents an analysis of EEG signals to detect true emotion/pretension: first by extracting the neural connectivity among selected brain lobes during arousal and manifestation of a true emotion, and then by testing whether the connectivity among the lobes are maintained while encountering an emotional context. In case the connectivity is manifested, the arousal of emotion is regarded as true emotion, otherwise it is considered as a pretension. Experimental results confirm that for positive emotions, the decoding accuracy of true (false) emotions is as high as 88% (72%), while for negative emotions, the classification accuracy falls off by a 12% margin for true emotions and 8% margin for false emotions. The proposed method has wide-spread applications to detect criminals, frauds and anti-socials.
Top1. Introduction
Emotion represents a psychological state of the human mind. Philosophers are of the view that emotion originates because of substantial changes in one’s personal situations (Ben-Ze'ev, 2002). . Arousal of emotion is often concurred with its manifestation on facial expressions, voice, gestures, postures, and physiological parameters, including body temperature, pulse rate, heart rate variability and cortical signal characteristics. Recognition of emotion from its facial and/or bodily manifestation thus is an interesting problem. Significant progress in recognition of emotion from its manifestation has been reported in the literature (Orozco, Rudovic, Gonzàlez, & Pantic,2014), (Lee & Narayanan, 2005), (Mitra & Acharya, 2007), (Murugappan, Rizon, Nagarajan & Yaacob, 2010). A survey of the literature reveals recognition accuracy of emotion depends largely on the modality of recognition (Chakraborty & Konar, 2009). For example, facial expressions and/or voice provide high classification accuracy of subjective emotion, while physiological parameters seem to have produced relatively less classification accuracy
Although manifestation of emotion is a good measure of the degree and class of emotion, sometimes the results of classification do not conform to the originated emotion, particularly for two reasons. First, the subject is not capable to develop manifestation of the aroused emotion. Secondly (and more importantly), the subject is unwilling to express her emotion on facial expression, voice and/or bodily gestures. Recognition of true emotion, however, is very important for the next-generation Human-Computer Interactive (HCI) systems. The present chapter will serve as a useful resource to discriminate true emotion from pretension. Experiments undertaken by previous researchers reveal that facial and bodily manifestation requires several 10 seconds after the onset of emotion arousal. However, cortical signals respond almost instantaneously at the onset of emotion-arousal. Naturally, voluntarily control of cortical signals to suppress or present false emotion while experiencing an emotion is difficult.
The chapter provides a thorough analysis of cortical signals to distinguish (classify) true emotions and pretensions. Early research on cortical signal analysis was restricted in good hospitals over the last two decades, particularly because of the excessive cost of EEG machines. Fortunately, EEG machines are now commercially available at a low price, and most of the research laboratories on brain science can afford it. EEG analysis on subjects in a given age-group of 24±4 years thus is performed to classify true emotion and pretension. The underlying principles to classify a given emotion into one of two classes: true emotion and pretension are briefly presented below.
An analysis of the brain map for emotion arousal reveals that emotion usually originates in the pre-frontal cortex and neural impulses associated with arousal of emotion pass to the pons region through motor cortex (Purves, Cabeza, Huettel, LaBar, Platt &Woldorff 2013). The motor cortex impulses activate muscles to generate electromyogram (EMG) signals, which in turn causes changes in facial expressions/gestures /postures for a given emotion (Morecraft, Stilwell-Morecraft & Rossing 2004), (Meister, Boroojerdi, Foltys, Sparing, Huber & Töpper, 2003.). The biological basis of manifestation on facial expression during arousal of an emotion is used in this paper to classify true emotion and pretension.
The computational model used here aims at determining the neuronal weights (connectivity) involved in mapping pre-frontal to motor cortex EEG features, next motor cortex EEG features to facial features and lastly, facial features back to pre-frontal EEG features during arousal of a given emotion. The training instances are collected for true emotion only in all circumstances. Naturally, after the weights are determined, the trained neural network is used to examine the closeness of the measured EEG features from the pre-frontal region and the predicted EEG features of the pre-frontal region. If the Euclidean norm of the measure of closeness is below a experimentally obtained prescribed threshold, the predicted emotion is declared as true emotion, else we call it a pretension.
Key Terms in this Chapter
Motor-Cortex: Motor cortex is the part of brain associated with planning, control and execution of movements. It forms the rear of the frontal lobe, located at the junction of frontal and parietal lobe. It is adjacent to somatosensory cortex which is a part of the parietal lobe.
Pretension: The voluntary suppression or manipulation of physical manifestations, representing the true psychological state of the human mind, is called pretension. Pretention can be detected by the study on non-voluntary manifestation in physiological responses including cardiovascular activity, skin conductance, respiration and electroencephalogram.
Human-Computer Interaction (HCI): HCI involves study, planning, design and uses of the interaction between humans and computers. It is a multidisciplinary avenue of behavioral sciences, computer science, media studies and design.
Negative Emotion: Emotions associated with negative valence are categorized as negative emotions. These include fear, disgust and anger.
Brain-Computer Interface (BCI): BCI is a communication pathway connecting the brain and computer. BCI, also known as brain machine interface (BMI), is often directed at creating assisting, augmenting and repairing human cognitive and sensory motor functions.
Positive Emotion: Emotions associated with positive valence (affective dimension indicative of pleasure) are categorized as positive emotions. These include happiness, relaxation and excitement.
Electroencephalogram (EEG): EEG is a recording of the brain’s spontaneous electrical activity. It is measured through electrodes placed on the scalp. These electrodes capture electric field resulting from the interchange of ions among the neurons in the brain.
Pre-Frontal Cortex: The pre-frontal cortex is an area covering the front part of the frontal lobe. It is associated with processing of cognitive functions including memory, language and emotions.