Recognizing facial expressions is a challenging task both for computers and humans. Although recent deep learning-based approaches are achieving high accuracy results in this task, research in this area is mainly focused on improving results using a single dataset for training and testing. This approach lacks generality when applied to new images or when using it in in-the-wild contexts due to diversity in humans (e.g., age, ethnicity) and differences in capture conditions (e.g., lighting or background). The cross-datasets approach can overcome these limitations. In this work we present a method to combine multiple datasets and we conduct an exhaustive evaluation of a proposed system based on a CNN analyzing and comparing performance using single and cross-dataset approaches with other architectures. Results using the proposed system ranged from 31.56% to 61.78% when used in a single-dataset approach with different well-known datasets and improved up to 73.05% when using a cross-dataset approach. Finally, to study the system and humans’ performance in facial expressions classification, we compare the results of 253 participants with the system. Results show an 83.53% accuracy for humans and a correlation exists between the results obtained by the participants and the CNN.

In this work an affective computing approach is used to study the human-robot interaction using a social robot to validate facial expressions in the wild. Our global goal is to evaluate that a social robot can be used to interact in a convincing manner with human users to recognize their potential emotions through facial expressions, contextual cues and bio-signals. In particular, this work is focused on analyzing facial expression. A social robot is used to validate a pre-trained convolutional neural network (CNN) which recognizes facial expressions. Facial expression recognition plays an important role in recognizing and understanding human emotion by robots. Robots equipped with expression recognition capabilities can also be a useful tool to get feedback from the users. The designed experiment allows evaluating a trained neural network in facial expressions using a social robot in a real environment. In this paper a comparison between the CNN accuracy and human experts is performed, in addition to analyze the interaction, attention and difficulty to perform a particular expression by 29 non-expert users. In the experiment, the robot leads the users to perform different facial expressions in motivating and entertaining way. At the end of the experiment, the users are quizzed about their experience with the robot. Finally, a set of experts and the CNN classify the expressions. The obtained results allow affirming that the use of social robot is an adequate interaction paradigm for the evaluation on facial expression.

Technology and videogames have been proven as motivating tools for working attention and complex communication skills, especially in children with autism spectrum disorder (ASD). In this work, we present two experiences that used interactive games for promoting communication and attention. The first game considers emotions in order to measure children’s attention, concentration and satisfaction, while the second uses tangible tabletops for fostering cognitive planning. The analysis of the results obtained allows to propose a new study integrating both, in which the tangible interactive game is complemented with the emotional trainer in a way that allows identifying and classifying children’s emotion with ASD when they collaborate to solve cognitively significant and contextualized challenges. The first application proposed is an emotional trainer application in which the child can work out the seven basic emotions (happiness, sadness, fear, disgust, anger, surprise and neutral). Further, a serious videogame is proposed: a 3D maze where the emotions can be captured. The second case study was carried out in a Special Education Center, where a set of activities for working cognitive planning was proposed. In this case, a tangible interactive tabletop was used to analyze, in students with ASD, how the communication processes with these interfaces affect to the attention, memory, successive and simultaneous processing that compose cognitive planning from the PASS model. The results of the first study, suggest that the autistic children did not act with previous planning, but they used their perception to adjust their actions a posteriori (that explains the higher number of collisions). On the second case study, the successive processing was not explored. The inclusion of the mazes of case study 1 to a semantic rich scenario could allow us to measure the prior planning and the emotions involved in the maze game. The new physiological sensors will also help to validate the emotions felt by the children. The first study has as objective the capability to imitate emotions and resolve a maze without semantic context. The second study organized all the actions from a semantic context close to users. The attention results presented by the second study are coherent with the first study and complement it showing that attention can be receptive or selective. In the first study case, the receptive attention was the focus of analysis. In the second case, both contributed to explain and understand how it can be developed from a videogame.

Facial expression recognition is vital for human behavior analysis, and deep learning has enabled models that can outperform humans. However, it is unclear how closely they mimic human processing. This study aims to explore the similarity between deep neural networks and human perception by comparing twelve different networks, including both general object classifiers and FER-specific models. We employ an innovative global explainable AI method to generate heatmaps, revealing crucial facial regions for the twelve networks trained on six facial expressions. We assess these results both quantitatively and qualitatively, comparing them to ground truth masks based on Friesen and Ekman’s description and among them. We use Intersection over Union (IoU) and normalized correlation coefficients for comparisons. We generate 72 heatmaps to highlight critical regions for each expression and architecture. Qualitatively, models with pre-trained weights show more similarity in heatmaps compared to those without pre-training. Specifically, eye and nose areas influence certain facial expressions, while the mouth is consistently important across all models and expressions. Quantitatively, we find low average IoU values (avg. 0.2702) across all expressions and architectures. The best-performing architecture averages 0.3269, while the worst-performing one averages 0.2066. Dendrograms, built with the normalized correlation coefficient, reveal two main clusters for most expressions: models with pre-training and models without pre-training. Findings suggest limited alignment between human and AI facial expression recognition, with network architectures influencing the similarity, as similar architectures prioritize similar facial regions.

Facial expression recognition plays an important role in human behaviour, communication, and interaction. Recent neural networks have demonstrated to perform well at its automatic recognition, with different explainability techniques available to make them more transparent. In this work, we propose a facial expression recognition study for people with intellectual disabilities that would be integrated into a social robot. We train two well-known neural networks with five databases of facial expressions and test them with two databases containing people with and without intellectual disabilities. Finally, we study in which regions the models focus to perceive a particular expression using two different explainability techniques: LIME and RISE, assessing the differences when used on images containing disabled and non-disabled people.