The Effects of Facial Attractiveness and Familiarity on Facial Expression Recognition

Scoring facial attractiveness with deep convolutional neural networks: How training on standardized images reduces the bias of facial expressions

OBJECTIVE

In many medical disciplines, facial attractiveness is part of the diagnosis, yet its scoring might be confounded by facial expressions. The intent was to apply deep convolutional neural networks (CNN) to identify how facial expressions affect facial attractiveness and to explore whether a dedicated training of the CNN is able to reduce the bias of facial expressions.

MATERIALS AND METHODS

Frontal facial images (n = 840) of 40 female participants (mean age 24.5 years) were taken adapting a neutral facial expression and the six universal facial expressions. Facial attractiveness was computed by means of a face detector, deep convolutional neural networks, standard support vector regression for facial beauty, visual regularized collaborative filtering and a regression technique for handling visual queries without rating history. CNN was first trained on random facial photographs from a dating website and then further trained on the Chicago Face Database (CFD) to increase its suitability to medical conditions. Both algorithms scored every image for attractiveness.

RESULTS

Facial expressions affect facial attractiveness scores significantly. Scores from CNN additionally trained on CFD had less variability between the expressions (range 54.3-60.9 compared to range: 32.6-49.5) and less variance within the scores (P ≤ .05), but also caused a shift in the ranking of the expressions' facial attractiveness.

CONCLUSION

Facial expressions confound attractiveness scores. Training on norming images generated scores less susceptible to distortion, but more difficult to interpret. Scoring facial attractiveness based on CNN seems promising, but AI solutions must be developed on CNN trained to recognize facial expressions as distractors.

Validation of the Edited Tromsø Infant Faces Database (E-TIF): A study on differences in the processing of children's emotional expressions

Images of emotional facial expressions are often used in emotion research, which has promoted the development of different databases. However, most of these standardized sets of images do not include images from infants under 2 years of age, which is relevant for psychology research, especially for perinatal psychology. The present study aims to validate the edited version of the Tromsø Infant Faces Database (E-TIF) in a large sample of participants. The original set of 119 pictures was edited. The pictures were cropped to remove nonrelevant information, fitted in an oval window, and converted to grayscale. Four hundred and eighty participants (72.9% women) took part in the study, rating the images on five dimensions: depicted emotion, clarity, intensity, valence, and genuineness. Valence scores were useful for discriminating between positive, negative, and neutral facial expressions. Results revealed that women were more accurate at recognizing emotions in children. Regarding parental status, parents, in comparison with nonparents, rated neutral expressions as more intense and genuine. They also rated sad, angry, disgusted, and fearful faces as less negative, and happy expressions as less positive. The editing and validation of the E-TIF database offers a useful tool for basic and experimental research in psychology.

Perception of Nasal Aesthetics: Nose or Face?

BACKGROUND

Aesthetic assessment of the nose might not be limited to the nose as an isolated facial unit but might be the result of a broader perception of the whole facial image. The aim of this study was to investigate if the aesthetic evaluation of a nose is made by sole observation of the nose or influenced by the other features of the face.

METHODS

Nose and face photographs of 20 voluntary models were taken from five different angles and merged into a single image of just the noses and whole faces of each model. First the nose and then the face photos were sent to a reviewer group consisting of 100 individuals from different professions to evaluate the nose's beauty on standard questionnaires with a four-week interval to complicate the recall process.

RESULTS

Aesthetic evaluation of the isolated nose (Group 1) and the nose with the whole face (Group 2) were compared through Visual Analog Scale (VAS) scores ranging from 1 to 10 (1: the least, 10: the highest). The mean VAS score of the Whole Face Group (5.26 ± 1.28) was significantly higher than the score of the Isolated Nose Group (4.50 ± 1.32) (p 0.001). There was no significant difference between the scores of reviewers considering their gender, profession, or experience of having an aesthetic operation previously.

CONCLUSION

Current study revealed that in the evaluation of the beauty of the nose, the holistic perception of the face is as important as the perfection of the angles and proportions within the nose itself.

LEVEL OF EVIDENCE V

This journal requires that authors assign a level of evidence to each article. For a full description of these Evidence-Based Medicine ratings, please refer to the Table of Contents or the online Instructions to Authors www.springer.com/00266 .

The effect of facial attractiveness on micro-expression recognition

Micro-expression (ME) is an extremely quick and uncontrollable facial movement that lasts for 40–200 ms and reveals thoughts and feelings that an individual attempts to cover up. Though much more difficult to detect and recognize, ME recognition is similar to macro-expression recognition in that it is influenced by facial features. Previous studies suggested that facial attractiveness could influence facial expression recognition processing. However, it remains unclear whether facial attractiveness could also influence ME recognition. Addressing this issue, this study tested 38 participants with two ME recognition tasks in a static condition or dynamically. Three different MEs (positive, neutral, and negative) at two attractiveness levels (attractive, unattractive). The results showed that participants recognized MEs on attractive faces much quicker than on unattractive ones, and there was a significant interaction between ME and facial attractiveness. Furthermore, attractive happy faces were recognized faster in both the static and the dynamic conditions, highlighting the happiness superiority effect. Therefore, our results provided the first evidence that facial attractiveness could influence ME recognition in a static condition or dynamically.

Neural responses to facial attractiveness in the judgments of moral goodness and moral beauty

The judgments of moral goodness and moral beauty objectively refer to the perception and evaluation of moral traits, which are generally influenced by facial attractiveness. For centuries, people have equated beauty with the possession of positive qualities, but it is not clear whether the association between beauty and positive qualities exerts a similarly implicit influence on people's responses to moral goodness and moral beauty, how it affects those responses, and what is the neural basis for such an effect. The present study is the first to examine the neural responses to facial attractiveness in the judgments of moral goodness and moral beauty. We found that beautiful faces in both moral judgments activated the left ventral occipitotemporal cortices sensitive to the geometric configuration of the faces, demonstrating that both moral goodness and moral beauty required the automatic visual analysis of geometrical configuration of attractive faces. In addition, compared to beautiful faces during moral goodness judgment, beautiful faces during moral beauty judgment induced unique activity in the ventral medial prefrontal cortex and midline cortical structures involved in the emotional-valenced information about attractive faces. The opposite comparison elicited specific activity in the left superior temporal cortex and premotor area, which play a critical role in the recognition of facial identity. Our results demonstrated that the neural responses to facial attractiveness in the process of higher order moral decision-makings exhibit both task-general and task-specific characteristics. Our findings contribute to the understanding of the essence of the relationship between morality and aesthetics.

In the mood to be social: Affective state influences facial emotion recognition in healthy adults

The ability to accurately recognize facial expressions is a key element of social interaction. Facial emotion recognition (FER) assessments show promise as a clinical screening and therapeutic tool, but realizing this potential requires better understanding of the stability of this skill. Transient mood states are known to bias emotion recognition in some contexts and may represent a critical factor impacting FER ability. In particular, it is unclear how natural fluctuations in individuals' mood state over time contribute to specific changes in the ability to recognize facial expressions. The current study tested 55 neurotypical participants across multiple visits using the Emotion Recognition test and found that fluctuations in positive and negative mood state altered recognition of specific emotions. Surprisingly, effects of mood state on emotion recognition were noncongruent; increased positive mood was associated with improved recognition of scared expressions but worsened recognition of happy expressions. Our results suggest that minor fluctuations in mood state in a neurotypical population affect emotion recognition. Therefore, mood should be taken into account by researchers and clinicians assessing FER skills. (PsycInfo Database Record (c) 2021 APA, all rights reserved).

Contrast Effect of Facial Attractiveness in Groups

Research on facial attractiveness is an important part of aesthetics. Most relevant studies in the area have focused on the influence of individual perspectives on facial attractiveness, but it is necessary to consider the effect of contextual information on facial attractiveness. In this study, we examine the influence on attractiveness of special faces in a given group. We define a “special face” as one that is significantly different from other members of the same group in terms of facial attractiveness. We conducted three experiments to explore the influence of different modes of presentation and central positions in a group on the judgment of attractiveness of the special face. The results show the following: (1) When the special face was part of a given group, the subjects made more extreme judgments than without it: that is, they judged the most attractive face as more attractive and the least as less attractive than when faces were presented alone. (2) The subjects rated the most attractive faces lower and the least attractive faces higher when the target faces in the middle of the group than in other positions. The results favored the contrast effect: when the subjects judged the attractiveness of target stimulus, they always compared it with the environment, which then became a reference in this regard. Moreover, the greater the amount of contextual information perceived, the higher the likelihood that assimilation would occur.

Research on a Noise Reduction Method Based on Multi-Resolution Singular Value Decomposition

Reducing noise pollution in signals is of great significance in the field of signal detection. In order to reduce the noise in the signal and improve the signal-to-noise ratio (SNR), this paper takes the singular value decomposition theory as the starting point, and constructs various singular value decomposition denoising models with multiple multi-division structures based on the two-division recursion singular value decomposition, and conducts a noise reduction analysis on two experimental signals containing noise of different power. Finally, the SNR and mean square error (MSE) are used as indicators to evaluate the noise reduction effect, it is verified that the two-division recursion singular value decomposition is the optimal noise reduction model. This noise reduction model is then applied to the diagnosis of faulty bearings. By this method, the fault signal is decomposed to reduce noise and the detail signal with maximum kurtosis is extracted for envelope spectrum analysis. Comparison of several traditional signal processing methods such as empirical modal decomposition (EMD), ensemble empirical mode decomposition (EEMD), variational mode decomposition (VMD), wavelet decomposition, etc. The results show that multi-resolution singular value decomposition (MRSVD) has better noise reduction effect and can effectively diagnose faulty bearings. This method is promising and has a good application prospect.