Human facial emotion detection is one of the challenging tasks in computer vision. Owing to high inter-class variance, it is hard for machine learning models to predict facial emotions accurately. Moreover, a person with several facial emotions increases the diversity and complexity of classification problems. In this paper, we have proposed a novel and intelligent approach for the classification of human facial emotions. The proposed approach comprises customized ResNet18 by employing transfer learning with the integration of triplet loss function (TLF), followed by SVM classification model. Using deep features from a customized ResNet18 trained with triplet loss, the proposed pipeline consists of a face detector used to locate and refine the face bounding box and a classifier to identify the facial expression class of discovered faces. RetinaFace is used to extract the identified face areas from the source image, and a ResNet18 model is trained on cropped face images with triplet loss to retrieve those features. An SVM classifier is used to categorize the facial expression based on the acquired deep characteristics. In this paper, we have proposed a method that can achieve better performance than state-of-the-art (SoTA) methods on JAFFE and MMI datasets. The technique is based on the triplet loss function to generate deep input image features. The proposed method performed well on the JAFFE and MMI datasets with an accuracy of 98.44% and 99.02%, respectively, on seven emotions; meanwhile, the performance of the method needs to be fine-tuned for the FER2013 and AFFECTNET datasets. Full article

Driver monitoring systems play an important role in lower to mid-level autonomous vehicles. Our work focuses on the detection of cognitive load as a component of driver-state estimation to improve traffic safety. By inducing single and dual-task workloads of increasing intensity on 51 subjects, while continuously measuring signals from multiple modalities, based on physiological measurements such as ECG, EDA, EMG, PPG, respiration rate, skin temperature and eye tracker data, as well as behavioral measurements such as action units extracted from facial videos, performance metrics like reaction time and subjective feedback using questionnaires, we create ADABase (Autonomous Driving Cognitive Load Assessment Database) As a reference method to induce cognitive load onto subjects, we use the well-established n-back test, in addition to our novel simulator-based k-drive test, motivated by real-world semi-autonomously vehicles. We extract expert features of all measurements and find significant changes in multiple modalities. Ultimately we train and evaluate machine learning algorithms using single and multimodal inputs to distinguish cognitive load levels. We carefully evaluate model behavior and study feature importance. In summary, we introduce a novel cognitive load test, create a cognitive load database, validate changes using statistical tests, introduce novel classification and regression tasks for machine learning and train and evaluate machine learning models. Full article

The monitoring of emotional state is important in the prevention and management of mental health problems and is increasingly being used to support affective computing. As such, researchers are exploring various modalities from which emotion can be inferred, such as through facial images or via electroencephalography (EEG) signals. Current research commonly investigates the performance of machine-learning-based emotion recognition systems by exposing users to stimuli that are assumed to elicit a single unchanging emotional response. Moreover, in order to demonstrate better results, many models are tested in evaluation frameworks that do not reflect realistic real-world implementations. Consequently, in this paper, we explore the design of EEG-based emotion recognition systems using longer, variable stimuli using the publicly available AMIGOS dataset. Feature engineering and selection results are evaluated across four different cross-validation frameworks, including versions of leave-one-movie-out (testing with a known user, but a previously unseen movie), leave-one-person-out (testing with a known movie, but a previously unseen person), and leave-one-person-and-movie-out (testing on both a new user and new movie). Results of feature selection lead to a 13% absolute improvement over comparable previously reported studies, and demonstrate the importance of evaluation framework on the design and performance of EEG-based emotion recognition systems. Full article

Rapid advancements in the medical field have drawn much attention to automatic emotion classification from EEG data. People’s emotional states are crucial factors in how they behave and interact physiologically. The diagnosis of patients’ mental disorders is one potential medical use. When feeling well, people work and communicate more effectively. Negative emotions can be detrimental to both physical and mental health. Many earlier studies that investigated the use of the electroencephalogram (EEG) for emotion classification have focused on collecting data from the whole brain because of the rapidly developing science of machine learning. However, researchers cannot understand how various emotional states and EEG traits are related. This work seeks to classify EEG signals’ positive, negative, and neutral emotional states by using a stacking-ensemble-based classification model that boosts accuracy to increase the efficacy of emotion classification using EEG. The selected features are used to train a model that was created using a random forest, light gradient boosting machine, and gradient-boosting-based stacking ensemble classifier (RLGB-SE), where the base classifiers random forest (RF), light gradient boosting machine (LightGBM), and gradient boosting classifier (GBC) were used at level 0. The meta classifier (RF) at level 1 is trained using the results from each base classifier to acquire the final predictions. The suggested ensemble model achieves a greater classification accuracy of 99.55%. Additionally, while comparing performance indices, the suggested technique outperforms as compared with the base classifiers. Comparing the proposed stacking strategy to state-of-the-art techniques, it can be seen that the performance for emotion categorization is promising. Full article

Vocal emotion recognition (VER) in natural speech, often referred to as speech emotion recognition (SER), remains challenging for both humans and computers. Applied fields including clinical diagnosis and intervention, social interaction research or Human Computer Interaction (HCI) increasingly benefit from efficient VER algorithms. Several feature sets were used with machine-learning (ML) algorithms for discrete emotion classification. However, there is no consensus for which low-level-descriptors and classifiers are optimal. Therefore, we aimed to compare the performance of machine-learning algorithms with several different feature sets. Concretely, seven ML algorithms were compared on the Berlin Database of Emotional Speech: Multilayer Perceptron Neural Network (MLP), J48 Decision Tree (DT), Support Vector Machine with Sequential Minimal Optimization (SMO), Random Forest (RF), k-Nearest Neighbor (KNN), Simple Logistic Regression (LOG) and Multinomial Logistic Regression (MLR) with 10-fold cross validation using four openSMILE feature sets (i.e., IS-09, emobase, GeMAPS and eGeMAPS). Results indicated that SMO, MLP and LOG show better performance (reaching to 87.85%, 84.00% and 83.74% accuracies, respectively) compared to RF, DT, MLR and KNN (with minimum 73.46%, 53.08%, 70.65% and 58.69% accuracies, respectively). Overall, the emobase feature set performed best. We discuss the implications of these findings for applications in diagnosis, intervention or HCI. Full article

Understanding of the perception of emotions or affective states in humans is important to develop emotion-aware systems that work in realistic scenarios. In this paper, the perception of emotions in naturalistic human interaction (audio–visual data) is studied using perceptual evaluation. For this purpose, a naturalistic audio–visual emotion database collected from TV broadcasts such as soap-operas and movies, called the IIIT-H Audio–Visual Emotion (IIIT-H AVE) database, is used. The database consists of audio-alone, video-alone, and audio–visual data in English. Using data of all three modes, perceptual tests are conducted for four basic emotions (angry, happy, neutral, and sad) based on category labeling and for two dimensions, namely arousal (active or passive) and valence (positive or negative), based on dimensional labeling. The results indicated that the participants’ perception of emotions was remarkably different between the audio-alone, video-alone, and audio–video data. This finding emphasizes the importance of emotion-specific features compared to commonly used features in the development of emotion-aware systems. Full article

The neural correlates of intentional emotion transfer by the music performer are not well investigated as the present-day research mainly focuses on the assessment of emotions evoked by music. In this study, we aim to determine whether EEG connectivity patterns can reflect differences in information exchange during emotional playing. The EEG data were recorded while subjects were performing a simple piano score with contrasting emotional intentions and evaluated the subjectively experienced success of emotion transfer. The brain connectivity patterns were assessed from the EEG data using the Granger Causality approach. The effective connectivity was analyzed in different frequency bands—delta, theta, alpha, beta, and gamma. The features that (1) were able to discriminate between the neutral baseline and the emotional playing and (2) were shared across conditions, were used for further comparison. The low frequency bands—delta, theta, alpha—showed a limited number of connections (4 to 6) contributing to the discrimination between the emotional playing conditions. In contrast, a dense pattern of connections between regions that was able to discriminate between conditions (30 to 38) was observed in beta and gamma frequency ranges. The current study demonstrates that EEG-based connectivity in beta and gamma frequency ranges can effectively reflect the state of the networks involved in the emotional transfer through musical performance, whereas utility of the low frequency bands (delta, theta, alpha) remains questionable. Full article

In this paper, we tackle the problem of predicting the affective responses of movie viewers, based on the content of the movies. Current studies on this topic focus on video representation learning and fusion techniques to combine the extracted features for predicting affect. Yet, these typically, while ignoring the correlation between multiple modality inputs, ignore the correlation between temporal inputs (i.e., sequential features). To explore these correlations, a neural network architecture—namely AttendAffectNet (AAN)—uses the self-attention mechanism for predicting the emotions of movie viewers from different input modalities. Particularly, visual, audio, and text features are considered for predicting emotions (and expressed in terms of valence and arousal). We analyze three variants of our proposed AAN: Feature AAN, Temporal AAN, and Mixed AAN. The Feature AAN applies the self-attention mechanism in an innovative way on the features extracted from the different modalities (including video, audio, and movie subtitles) of a whole movie to, thereby, capture the relationships between them. The Temporal AAN takes the time domain of the movies and the sequential dependency of affective responses into account. In the Temporal AAN, self-attention is applied on the concatenated (multimodal) feature vectors representing different subsequent movie segments. In the Mixed AAN, we combine the strong points of the Feature AAN and the Temporal AAN, by applying self-attention first on vectors of features obtained from different modalities in each movie segment and then on the feature representations of all subsequent (temporal) movie segments. We extensively trained and validated our proposed AAN on both the MediaEval 2016 dataset for the Emotional Impact of Movies Task and the extended COGNIMUSE dataset. Our experiments demonstrate that audio features play a more influential role than those extracted from video and movie subtitles when predicting the emotions of movie viewers on these datasets. The models that use all visual, audio, and text features simultaneously as their inputs performed better than those using features extracted from each modality separately. In addition, the Feature AAN outperformed other AAN variants on the above-mentioned datasets, highlighting the importance of taking different features as context to one another when fusing them. The Feature AAN also performed better than the baseline models when predicting the valence dimension. Full article

Besides facial or gesture-based emotion recognition, Electroencephalogram (EEG) data have been drawing attention thanks to their capability in countering the effect of deceptive external expressions of humans, like faces or speeches. Emotion recognition based on EEG signals heavily relies on the features and their delineation, which requires the selection of feature categories converted from the raw signals and types of expressions that could display the intrinsic properties of an individual signal or a group of them. Moreover, the correlation or interaction among channels and frequency bands also contain crucial information for emotional state prediction, and it is commonly disregarded in conventional approaches. Therefore, in our method, the correlation between 32 channels and frequency bands were put into use to enhance the emotion prediction performance. The extracted features chosen from the time domain were arranged into feature-homogeneous matrices, with their positions following the corresponding electrodes placed on the scalp. Based on this 3D representation of EEG signals, the model must have the ability to learn the local and global patterns that describe the short and long-range relations of EEG channels, along with the embedded features. To deal with this problem, we proposed the 2D CNN with different kernel-size of convolutional layers assembled into a convolution block, combining features that were distributed in small and large regions. Ten-fold cross validation was conducted on the DEAP dataset to prove the effectiveness of our approach. We achieved the average accuracies of 98.27% and 98.36% for arousal and valence binary classification, respectively. Full article

Music videos contain a great deal of visual and acoustic information. Each information source within a music video influences the emotions conveyed through the audio and video, suggesting that only a multimodal approach is capable of achieving efficient affective computing. This paper presents an affective computing system that relies on music, video, and facial expression cues, making it useful for emotional analysis. We applied the audio–video information exchange and boosting methods to regularize the training process and reduced the computational costs by using a separable convolution strategy. In sum, our empirical findings are as follows: (1) Multimodal representations efficiently capture all acoustic and visual emotional clues included in each music video, (2) the computational cost of each neural network is significantly reduced by factorizing the standard 2D/3D convolution into separate channels and spatiotemporal interactions, and (3) information-sharing methods incorporated into multimodal representations are helpful in guiding individual information flow and boosting overall performance. We tested our findings across several unimodal and multimodal networks against various evaluation metrics and visual analyzers. Our best classifier attained 74% accuracy, an f1-score of 0.73, and an area under the curve score of 0.926. Full article

Decades of scientific research have been conducted on developing and evaluating methods for automated emotion recognition. With exponentially growing technology, there is a wide range of emerging applications that require emotional state recognition of the user. This paper investigates a robust approach for multimodal emotion recognition during a conversation. Three separate models for audio, video and text modalities are structured and fine-tuned on the MELD. In this paper, a transformer-based crossmodality fusion with the EmbraceNet architecture is employed to estimate the emotion. The proposed multimodal network architecture can achieve up to 65% accuracy, which significantly surpasses any of the unimodal models. We provide multiple evaluation techniques applied to our work to show that our model is robust and can even outperform the state-of-the-art models on the MELD. Full article

Emotion recognition plays an important role in human–computer interactions. Recent studies have focused on video emotion recognition in the wild and have run into difficulties related to occlusion, illumination, complex behavior over time, and auditory cues. State-of-the-art methods use multiple modalities, such as frame-level, spatiotemporal, and audio approaches. However, such methods have difficulties in exploiting long-term dependencies in temporal information, capturing contextual information, and integrating multi-modal information. In this paper, we introduce a multi-modal flexible system for video-based emotion recognition in the wild. Our system tracks and votes on significant faces corresponding to persons of interest in a video to classify seven basic emotions. The key contribution of this study is that it proposes the use of face feature extraction with context-aware and statistical information for emotion recognition. We also build two model architectures to effectively exploit long-term dependencies in temporal information with a temporal-pyramid model and a spatiotemporal model with “Conv2D+LSTM+3DCNN+Classify” architecture. Finally, we propose the best selection ensemble to improve the accuracy of multi-modal fusion. The best selection ensemble selects the best combination from spatiotemporal and temporal-pyramid models to achieve the best accuracy for classifying the seven basic emotions. In our experiment, we take benchmark measurement on the AFEW dataset with high accuracy. Full article

Recognizing user emotions while they watch short-form videos anytime and anywhere is essential for facilitating video content customization and personalization. However, most works either classify a single emotion per video stimuli, or are restricted to static, desktop environments. To address this, we propose a correlation-based emotion recognition algorithm (CorrNet) to recognize the valence and arousal (V-A) of each instance (fine-grained segment of signals) using only wearable, physiological signals (e.g., electrodermal activity, heart rate). CorrNet takes advantage of features both inside each instance (intra-modality features) and between different instances for the same video stimuli (correlation-based features). We first test our approach on an indoor-desktop affect dataset (CASE), and thereafter on an outdoor-mobile affect dataset (MERCA) which we collected using a smart wristband and wearable eyetracker. Results show that for subject-independent binary classification (high-low), CorrNet yields promising recognition accuracies: $76.37\%$ and $74.03\%$ for V-A on CASE, and $70.29\%$ and $68.15\%$ for V-A on MERCA. Our findings show: (1) instance segment lengths between 1–4 s result in highest recognition accuracies (2) accuracies between laboratory-grade and wearable sensors are comparable, even under low sampling rates (≤64 Hz) (3) large amounts of neutral V-A labels, an artifact of continuous affect annotation, result in varied recognition performance. Full article