Information For Emotion Recognition Research Articles

Eyes on Emotion: Dynamic Gaze Allocation During Emotion Perception From Speech-Like Stimuli.

The majority of emotional expressions used in daily communication are multimodal and dynamic in nature. Consequently, one would expect that human observers utilize specific perceptual strategies to process emotions and to handle the multimodal and dynamic nature of emotions. However, our present knowledge on these strategies is scarce, primarily because most studies on emotion perception have not fully covered this variation, and instead used static and/or unimodal stimuli with few emotion categories. To resolve this knowledge gap, the present study examined how dynamic emotional auditory and visual information is integrated into a unified percept. Since there is a broad spectrum of possible forms of integration, both eye movements and accuracy of emotion identification were evaluated while observers performed an emotion identification task in one of three conditions: audio-only, visual-only video, or audiovisual video. In terms of adaptations of perceptual strategies, eye movement results showed a shift in fixations toward the eyes and away from the nose and mouth when audio is added. Notably, in terms of task performance, audio-only performance was mostly significantly worse than video-only and audiovisual performances, but performance in the latter two conditions was often not different. These results suggest that individuals flexibly and momentarily adapt their perceptual strategies to changes in the available information for emotion recognition, and these changes can be comprehensively quantified with eye tracking.

Parallelized Convolutional Recurrent Neural Network With Spectral Features for Speech Emotion Recognition

Speech is the most effective way for people to exchange complex information. Recognition of emotional information contained in speech is one of the important challenges in the field of artificial intelligence. To better acquire emotional features in speech signals, a parallelized convolutional recurrent neural network (PCRN) with spectral features is proposed for speech emotion recognition. First, frame-level features are extracted from each utterance and, a long short-term memory is employed to learn these features frame by frame. At the same time, the deltas and delta-deltas of the log Mel-spectrogram are calculated and reconstructed into three channels (static, delta, and delta-delta); these 3-D features are learned by a convolutional neural network (CNN). Then, the two learned high-level features are fused and batch normalized. Finally, a SoftMax classifier is used to classify emotions. Our PCRN model simultaneously processes two different types of features in parallel to better learn the subtle changes in emotion. The experimental results on four public datasets show the superiority of our proposed method, which is better than the previous works.

Cultural influences in recognition of emotional information: A cultural neuroscience approach

Cultural influences in recognition of emotional information: A cultural neuroscience approach

A Survey of Affective Computing for Stress Detection: Evaluating technologies in stress detection for better health

As we become more aware of the connection between emotional states and physical health, affective computing continues to rise as a field of interest. Affective computing uses both hardware and software technology to detect the affective state of a person. It is an active research area that has seen much growth in technology geared toward affective state analysis. Its origin is credited to Dr. Rosalind Picard of the Massachusetts Institute of Technology (MIT) when she published her 1995 article on affective computing [1]. It has since become a modern branch of computer science for human-computer interfaces [2], [3]. This stem of computer science has two main veins: 1) detection and recognition of emotional information and 2) simulation of emotion in computational devices. The focus of the current survey is the detection and recognition of emotions as affective states.

Severity of depressive symptoms and oscillatory responses to emotional facial expressions

We studies the differences in oscillatory responses to emotional facial expressions in 40 subjects with different severity of depressive symptoms (19 men and 21 women) aged 18-30 years. In subjects with low severity of depressive symptoms (group 2), the perception of happy facial expressions was marked by lower delta synchronization as compared with angry and neutral expressions; in subjects with high severity of depressive chronization as compared with angry and neutral expression; in subjects with high severity of depressive symptoms (group 1), the perception of happy faces was marked by higher delta synchronization. Since an increase in delta oscillations is usually observed in aversive conditions, we suggest that happy facial expressions are perceived as negative stimuli by the subjects of group 1. The perception of angry facial expressions was accompanied by alpha band desynchronization in Group 2 and by alpha synchronization in Group 1. Basing on Klimesch's theory, this effect suggests that the subjects of group 1 are initially set up for the perception of negative emotional information. The effect of emotional stimulus category was significant in group 2, but not significant in group 1, which is an evidence of disorders in the recognition of emotional information in depression-prone individuals.

Automatic integration of social information in emotion recognition.

This study investigated the automaticity of the influence of social inference on emotion recognition. Participants were asked to recognize dynamic facial expressions of emotion (fear or anger in Experiment 1 and blends of fear and surprise or of anger and disgust in Experiment 2) in a target face presented at the center of a screen while a subliminal contextual face appearing in the periphery expressed an emotion (fear or anger) or not (neutral) and either looked at the target face or not. Results of Experiment 1 revealed that recognition of the target emotion of fear was improved when a subliminal angry contextual face gazed toward-rather than away from-the fearful face. We replicated this effect in Experiment 2, in which facial expression blends of fear and surprise were more often and more rapidly categorized as expressing fear when the subliminal contextual face expressed anger and gazed toward-rather than away from-the target face. With the contextual face appearing for 30 ms in total, including only 10 ms of emotion expression, and being immediately masked, our data provide the first evidence that social influence on emotion recognition can occur automatically.

Supplemental Material for Automatic Integration of Social Information in Emotion Recognition

Supplemental Material for Automatic Integration of Social Information in Emotion Recognition

Emotional Verbal Learning Test: Development and Psychometric Properties

Memory deficits are a common feature of neurological and psychiatric disorders. Measures designed to evaluate memory in clinical populations have distinguished between memory for verbal and visual information; however, few tests assess the recall and recognition of emotional information, despite evidence suggesting that brain regions are differentially involved in memory for emotional and neutral stimuli and that affective disturbances are common in psychiatric and neurological disorders. The present study reports the test development and psychometric properties of the Emotional Verbal Learning Test (EVLT), a new neuropsychological measure that allows for the examination of emotional learning and memory. Psychometric analyses indicated that the EVLT has good internal consistency and test-retest reliability, as well as discriminant validity, clinical utility, and sensitivity to mood-congruency effects. This new measure has potential to be a valuable research and a clinical tool in the assessment of emotional memory and learning in healthy individuals and persons with neuropsychiatric disorders.

Facial Emotional Classifier For Natural Interaction

The recognition of emotional information is a key step toward giving computers the ability to interact more naturally and intelligently with people. We present a simple and computationally feasible method to perform automatic emotional classification of facial expressions. We propose the use of a set of characteristic facial points (that are part of the MPEG4 feature points) to extract relevant emotional information (basically five distances, presence of wrinkles in the eyebrow and mouth shape). The method defines and detects the six basic emotions (plus the neutral one) in terms of this information and has been fine-tuned with a database of more than 1500 images. The system has been integrated in a 3D engine for managing virtual characters, allowing the exploration of new forms of natural interaction.

Evidence for distinct contributions of form and motion information to the recognition of emotions from body gestures

The importance of kinematics in emotion perception from body movement has been widely demonstrated. Evidence also suggests that the perception of biological motion relies to some extent on information about spatial and spatiotemporal form, yet the contribution of such form-related cues to emotion perception remains unclear. This study reports, for the first time, the relative effects on emotion recognition of inverting and motion-reversing patch-light compared to fully illuminated displays of whole-body emotion gestures. Inverting the gesture movies or playing them backwards significantly impaired emotion classification accuracy, but did so more for patch-light displays than for identical but fully illuminated movement sequences. This result suggests that inversion impairs the processing of form information related to the configuration of body parts, and reversal impairs the sequencing of form changes, more than these manipulations impair the processing of kinematic cues. This effect was strongest for inversion, suggesting an important role for configural information in emotion recognition. Nevertheless, even in combination these stimulus manipulations did not abolish above chance recognition of any of the emotions, suggesting that kinematics help distinguish emotions expressed by body gestures. Disproportionate impairments in recognition accuracy were observed for fear and disgust under inversion, and for fear under motion reversal, suggesting a greater role for form-related cues in the perception of these emotions.