De plus en plus d’applications ont pour objectif d’automatiser l’analyse des comportements humains afin d’aider les experts qui réalisent actuellement ces analyses. Cette thèse traite de l’analyse des expressions faciales qui fournissent des informations clefs sur ces comportements.Les travaux réalisés portent sur une solution innovante, basée sur l’organisation des expressions, permettant de définir efficacement une expression d’un visage.Nous montrons que l’organisation des expressions, telle que définie, est universelle : une expression est alors caractérisée par son intensité et sa position relative par rapport aux autres expressions. La solution est comparée aux méthodes classiques et montre une augmentation significative des résultats de reconnaissance sur 14 expressions non basiques. La méthode a été étendue à des sujets inconnus. L’idée principale est de créer un espace d’apparence plausible spécifique à la personne inconnue en synthétisant ses expressions basiques à partir de déformations apprises sur d’autres sujets et appliquées sur le neutre du sujet inconnu. La solution est aussi mise à l’épreuve dans un environnement multimodal dont l’objectif est la reconnaissance d’émotions lors de conversations spontanées. Notre méthode a été mise en œuvre dans le cadre du challenge international AVEC 2012 (Audio/Visual Emotion Challenge) où nous avons fini 2nd, avec des taux de reconnaissance très proches de ceux obtenus par les vainqueurs. La comparaison des deux méthodes (la nôtre et celles des vainqueurs) semble montrer que l’extraction des caractéristiques pertinentes est la clef de tels systèmes. / More and more applications aim at automating the analysis of human behavior to assist or replace the experts who are conducting these analyzes. This thesis deals with the analysis of facial expressions, which provide key information on these behaviors.Our work proposes an innovative solution to effectively define a facial expression, regardless of the morphology of the subject. The approach is based on the organization of expressions.We show that the organization of expressions, such as defined, is universal and can be effectively used to uniquely define an expression. One expression is given by its intensity and its relative position to the other expressions. The solution is compared with the conventional methods based on appearance data and shows a significant increase in recognition results of 14 non-basic expressions. The method has been extended to unknown subjects. The main idea is to create a plausible appearance space dedicated to the unknown person by synthesizing its basic expressions from deformations learned on other subjects and applied to the neutral face of the unknown subject. The solution is tested in a more comprehensive multimodal environment, whose aim is the recognition of emotions in spontaneous conversations. Our method has been implemented in the international challenge AVEC 2012 (Audio / Visual Emotion Challenge) where we finished 2nd, with recognition rates very close to the winners’ ones. Comparison of both methods (ours and the winners’ one) seems to show that the extraction of relevant features is the key to such systems.

This thesis studies a specific population of children exposed to domestic violence. The theoretical part is focused on the impacts of domestic violence on emotionality and emotional development of children. It also deals with the problematic aspects of children's emotionality that are believed to be disturbed by this traumatizing experience. These particular areas were chosen based on present studies focused on the children exposed to domestic violence. The empirical part of the study was developed in cooperation with the Locika centre. It studies chosen aspects of children's emotionality in a quantitative and qualitative way. Particularly, it investigates the ability to distinguish emotions and the ability of emotion regulation within the overall cognitive development.

Master of Science / Department of Mechanical and Nuclear Engineering / Akira T. Tokuhiro / Robots will eventually become common everyday items. However before this becomes a reality, robots would need to learn be socially interactive. Since humans communicate much more information through expression than through actual spoken words, expression recognition is an important aspect in the development of social robots. Automatic recognition of emotional expressions has a number of potential applications other than just social robots. It can be used in systems that make sure the operator is alert at all times, or it can be used for psycho-analysis or cognitive studies. Emotional expressions are not always deliberate and can also occur without the person being aware of them. Recognizing these involuntary expressions provide an insight into the persons thought, state of mind and could be used as indicators for a hidden intent. In this research we developed an initial multi-modal emotion recognition system using cues from emotional expressions in face and voice. This is achieved by extracting features from each of the modalities using signal processing techniques, and then classifying these features with the help of artificial neural networks. The features extracted from the face are the eyes, eyebrows, mouth and nose; this is done using image processing techniques such as seeded region growing algorithm, particle swarm optimization and general properties of the feature being extracted. In contrast features of interest in speech are pitch, formant frequencies and mel spectrum along with some statistical properties such as mean and median and also the rate of change of these properties. These features are extracted using techniques such as Fourier transform and linear predictive coding. We have developed a toolbox that can read an audio and/or video file and perform emotion recognition on the face in the video and speech in the audio channel. The features extracted from the face and voices are independently classified into emotions using two separate feed forward type of artificial neural networks. This toolbox then presents the output of the artificial neural networks from one/both the modalities on a synchronized time scale. Some interesting results from this research is consistent misclassification of facial expressions between two databases, suggesting a cultural basis for this confusion. Addition of voice component has been shown to partially help in better classification.

No description available.

<div>This thesis consists of four sections which are related with four research projects.</div><div><br></div><div>The first section is about Page Classification. In this section, we extend our previous approach which could classify 3 classes of pages: Text, Picture and Mixed, to 5 classes which are: Text, Picture, Mixed, Receipt and Highlight. We first design new features to define those two new classes and then use DAG-SVM to classify those 5 classes of images. Based on the results, our algorithm performs well and is able to classify 5 types of pages.</div><div><br></div><div>The second section is about Fading Detection. In this section, we develop an algorithm that can automatically detect fading for both text and non-text region. For text region, we first do global alignment and then perform local alignment. After that, we create a 3D color node system, assign each connected component to a color node and get the color difference between raster page connected component and scanned page connected. For non-text region, after global alignment, we divide the page into "super pixels" and get the color difference between raster super pixels and testing super pixels. Compared with the traditional method that uses a diagnostic page, our method is more efficient and effective.</div><div><br></div><div>The third section is about CNN Based Emotion Recognition. In this section, we build our own emotion recognition classification and regression system from scratch. It includes data set collection, data preprocessing, model training and testing. We extend the model to real-time video application and it performs accurately and smoothly. We also try another approach of solving the emotion recognition problem using Facial Action Unit detection. By extracting Facial Land Mark features and adopting SVM training framework, the Facial Action Unit approach achieves comparable accuracy to the CNN based approach.</div><div><br></div><div>The forth section is about Saliency Based Image Quality Assessment and Cropping. In this section, we propose a method of doing image quality assessment and recomposition with the help of image saliency information. Saliency is the remarkable region of an image that attracts people's attention easily and naturally. By showing everyday examples as well as our experimental results, we demonstrate the fact that, utilizing the saliency information will be beneficial for both tasks.</div>

No description available.

Tesis por compendio / [ES] El uso de la realidad virtual (RV) se ha incrementado notablemente en la comunidad científica para la investigación del comportamiento humano. En particular, la RV inmersiva ha crecido debido a la democratización de las gafas de realidad virtual o head mounted displays (HMD), que ofrecen un alto rendimiento con una inversión económica. Uno de los campos que ha emergido con fuerza en la última década es el Affective Computing, que combina psicofisiología, informática, ingeniería biomédica e inteligencia artificial, desarrollando sistemas que puedan reconocer emociones automáticamente. Su progreso es especialmente importante en el campo de la investigación del comportamiento humano, debido al papel fundamental que las emociones juegan en muchos procesos psicológicos como la percepción, la toma de decisiones, la creatividad, la memoria y la interacción social. Muchos estudios se han centrado en intentar obtener una metodología fiable para evocar y automáticamente identificar estados emocionales, usando medidas fisiológicas objetivas y métodos de aprendizaje automático. Sin embargo, la mayoría de los estudios previos utilizan imágenes, audios o vídeos para generar los estados emocionales y, hasta donde llega nuestro conocimiento, ninguno de ellos ha desarrollado un sistema de reconocimiento emocional usando RV inmersiva. Aunque algunos trabajos anteriores sí analizan las respuestas fisiológicas en RV inmersivas, estos no presentan modelos de aprendizaje automático para procesamiento y clasificación automática de bioseñales. Además, un concepto crucial cuando se usa la RV en investigación del comportamiento humano es la validez: la capacidad de evocar respuestas similares en un entorno virtual a las evocadas por el espacio físico. Aunque algunos estudios previos han usado dimensiones psicológicas y cognitivas para comparar respuestas entre entornos reales y virtuales, las investigaciones que analizan respuestas fisiológicas o comportamentales están mucho menos extendidas. Según nuestros conocimientos, este es el primer trabajo que compara entornos físicos con su réplica en RV, empleando respuestas fisiológicas y algoritmos de aprendizaje automático y analizando la capacidad de la RV de transferir y extrapolar las conclusiones obtenidas al entorno real que se está simulando. El objetivo principal de la tesis es validar el uso de la RV inmersiva como una herramienta de estimulación emocional usando respuestas psicofisiológicas y comportamentales en combinación con algoritmos de aprendizaje automático, así como realizar una comparación directa entre un entorno real y virtual. Para ello, se ha desarrollado un protocolo experimental que incluye entornos emocionales 360º, un museo real y una virtualización 3D altamente realista del mismo museo. La tesis presenta novedosas contribuciones del uso de la RV inmersiva en la investigación del comportamiento humano, en particular en lo relativo al estudio de las emociones. Esta ayudará a aplicar metodologías a estímulos más realistas para evaluar entornos y situaciones de la vida diaria, superando las actuales limitaciones de la estimulación emocional que clásicamente ha incluido imágenes, audios o vídeos. Además, en ella se analiza la validez de la RV realizando una comparación directa usando una simulación altamente realista. Creemos que la RV inmersiva va a revolucionar los métodos de estimulación emocional en entornos de laboratorio. Además, su sinergia junto a las medidas fisiológicas y las técnicas de aprendizaje automático, impactarán transversalmente en muchas áreas de investigación como la arquitectura, la salud, la evaluación psicológica, el entrenamiento, la educación, la conducción o el marketing, abriendo un nuevo horizonte de oportunidades para la comunidad científica. La presente tesis espera contribuir a caminar en esa senda. / [EN] In recent years the scientific community has significantly increased its use of virtual reality (VR) technologies in human behaviour research. In particular, the use of immersive VR has grown due to the introduction of affordable, high performance head mounted displays (HMDs). Among the fields that has strongly emerged in the last decade is affective computing, which combines psychophysiology, computer science, biomedical engineering and artificial intelligence in the development of systems that can automatically recognize emotions. The progress of affective computing is especially important in human behaviour research due to the central role that emotions play in many background processes, such as perception, decision-making, creativity, memory and social interaction. Several studies have tried to develop a reliable methodology to evoke and automatically identify emotional states using objective physiological measures and machine learning methods. However, the majority of previous studies used images, audio or video to elicit emotional statements; to the best of our knowledge, no previous research has developed an emotion recognition system using immersive VR. Although some previous studies analysed physiological responses in immersive VR, they did not use machine learning techniques for biosignal processing and classification. Moreover, a crucial concept when using VR for human behaviour research is validity: the capacity to evoke a response from the user in a simulated environment similar to the response that might be evoked in a physical environment. Although some previous studies have used psychological and cognitive dimensions to compare responses in real and virtual environments, few have extended this research to analyse physiological or behavioural responses. Moreover, to our knowledge, this is the first study to compare VR scenarios with their real-world equivalents using physiological measures coupled with machine learning algorithms, and to analyse the ability of VR to transfer and extrapolate insights obtained from VR environments to real environments. The main objective of this thesis is, using psycho-physiological and behavioural responses in combination with machine learning methods, and by performing a direct comparison between a real and virtual environment, to validate immersive VR as an emotion elicitation tool. To do so we develop an experimental protocol involving emotional 360º environments, an art exhibition in a real museum, and a highly-realistic 3D virtualization of the same art exhibition. This thesis provides novel contributions to the use of immersive VR in human behaviour research, particularly in relation to emotions. VR can help in the application of methodologies designed to present more realistic stimuli in the assessment of daily-life environments and situations, thus overcoming the current limitations of affective elicitation, which classically uses images, audio and video. Moreover, it analyses the validity of VR by performing a direct comparison using highly-realistic simulation. We believe that immersive VR will revolutionize laboratory-based emotion elicitation methods. Moreover, its synergy with physiological measurement and machine learning techniques will impact transversely in many other research areas, such as architecture, health, assessment, training, education, driving and marketing, and thus open new opportunities for the scientific community. The present dissertation aims to contribute to this progress. / [CA] L'ús de la realitat virtual (RV) s'ha incrementat notablement en la comunitat científica per a la recerca del comportament humà. En particular, la RV immersiva ha crescut a causa de la democratització de les ulleres de realitat virtual o head mounted displays (HMD), que ofereixen un alt rendiment amb una reduïda inversió econòmica. Un dels camps que ha emergit amb força en l'última dècada és el Affective Computing, que combina psicofisiologia, informàtica, enginyeria biomèdica i intel·ligència artificial, desenvolupant sistemes que puguen reconéixer emocions automàticament. El seu progrés és especialment important en el camp de la recerca del comportament humà, a causa del paper fonamental que les emocions juguen en molts processos psicològics com la percepció, la presa de decisions, la creativitat, la memòria i la interacció social. Molts estudis s'han centrat en intentar obtenir una metodologia fiable per a evocar i automàticament identificar estats emocionals, utilitzant mesures fisiològiques objectives i mètodes d'aprenentatge automàtic. No obstant això, la major part dels estudis previs utilitzen imatges, àudios o vídeos per a generar els estats emocionals i, fins on arriba el nostre coneixement, cap d'ells ha desenvolupat un sistema de reconeixement emocional mitjançant l'ús de la RV immersiva. Encara que alguns treballs anteriors sí que analitzen les respostes fisiològiques en RV immersives, aquests no presenten models d'aprenentatge automàtic per a processament i classificació automàtica de biosenyals. A més, un concepte crucial quan s'utilitza la RV en la recerca del comportament humà és la validesa: la capacitat d'evocar respostes similars en un entorn virtual a les evocades per l'espai físic. Encara que alguns estudis previs han utilitzat dimensions psicològiques i cognitives per a comparar respostes entre entorns reals i virtuals, les recerques que analitzen respostes fisiològiques o comportamentals estan molt menys esteses. Segons els nostres coneixements, aquest és el primer treball que compara entorns físics amb la seua rèplica en RV, emprant respostes fisiològiques i algorismes d'aprenentatge automàtic i analitzant la capacitat de la RV de transferir i extrapolar les conclusions obtingudes a l'entorn real que s'està simulant. L'objectiu principal de la tesi és validar l'ús de la RV immersiva com una eina d'estimulació emocional usant respostes psicofisiològiques i comportamentals en combinació amb algorismes d'aprenentatge automàtic, així com realitzar una comparació directa entre un entorn real i virtual. Per a això, s'ha desenvolupat un protocol experimental que inclou entorns emocionals 360º, un museu real i una virtualització 3D altament realista del mateix museu. La tesi presenta noves contribucions de l'ús de la RV immersiva en la recerca del comportament humà, en particular quant a l'estudi de les emocions. Aquesta ajudarà a aplicar metodologies a estímuls més realistes per a avaluar entorns i situacions de la vida diària, superant les actuals limitacions de l'estimulació emocional que clàssicament ha inclòs imatges, àudios o vídeos. A més, en ella s'analitza la validesa de la RV realitzant una comparació directa usant una simulació altament realista. Creiem que la RV immersiva revolucionarà els mètodes d'estimulació emocional en entorns de laboratori. A més, la seua sinergia al costat de les mesures fisiològiques i les tècniques d'aprenentatge automàtic, impactaran transversalment en moltes àrees de recerca com l'arquitectura, la salut, l'avaluació psicològica, l'entrenament, l'educació, la conducció o el màrqueting, obrint un nou horitzó d'oportunitats per a la comunitat científica. La present tesi espera contribuir a caminar en aquesta senda. / Marín Morales, J. (2020). Modelling human emotions using immersive virtual reality, physiological signals and behavioural responses [Tesis doctoral no publicada]. Universitat Politècnica de València. https://doi.org/10.4995/Thesis/10251/148717 / TESIS / Compendio

No description available.

With the breakthrough of machine learning techniques, the research concerning music emotion classification has been getting notable progress combining various audio features and state-of-the-art machine learning models. Still, it is known that the way to preprocess music samples and to choose which machine classification algorithm to use depends on data sets and the objective of each project work. The collaborating company of this thesis, Ichigoichie AB, is currently developing a system to categorize music data into positive/negative classes. To enhance the accuracy of the existing system, this project aims to figure out the best model through experiments with six audio features (Mel spectrogram, MFCC, HPSS, Onset, CENS, Tonnetz) and several machine learning models including deep neural network models for the classification task. For each model, hyperparameter tuning is performed and the model evaluation is carried out according to pareto optimality with regard to accuracy and execution time. The results show that the most promising model accomplished 95% correct classification with an execution time of less than 15 seconds. / I och med genombrottet av maskininlärningstekniker har forskning kring känsloklassificering i musik sett betydande framsteg genom att kombinera olikamusikanalysverktyg med nya maskinlärningsmodeller. Trots detta är hur man förbehandlar ljuddatat och valet av vilken maskinklassificeringsalgoritm som ska tillämpas beroende på vilken typ av data man arbetar med samt målet med projektet. Denna uppsats samarbetspartner, Ichigoichie AB, utvecklar för närvarande ett system för att kategorisera musikdata enligt positiva och negativa känslor. För att höja systemets noggrannhet är målet med denna uppsats att experimentellt hitta bästa modellen baserat på sex musik-egenskaper (Mel-spektrogram, MFCC, HPSS, Onset, CENS samt Tonnetz) och ett antal olika maskininlärningsmodeller, inklusive Deep Learning-modeller. Varje modell hyperparameteroptimeras och utvärderas enligt paretooptimalitet med hänsyn till noggrannhet och beräkningstid. Resultaten visar att den mest lovande modellen uppnådde 95% korrekt klassificering med en beräkningstid på mindre än 15 sekunder.

Digital games are being increasingly implemented in the educational sector to improvevarious skills. Among these the effectiveness of digital game-based learning (DGBL) on the socio-emotional ability of individuals have been investigated with overall positive results. However, a limited number of studies have investigated the effectiveness of DGBL on this ability in children with intellectual disability and no studies have researched whether DGBL could be effective on the socio-emotional skills of children with this form of disability. Thus, the current study aimed at investigating the effectiveness of DGBL on the specific socio-emotional skill of emotion recognition in children with intellectual disability in the educational sector. The following research question was formulated: Does DGBL increase the socio-emotional skill of emotion recognition in children with intellectual disability. To answer this question a quasiexperimental one-group pretest-posttest design was adopted where participants engaged in a DGBL-intervention where they played a game aimed at improving their emotion recognition ability. Participants were selected via purposive sampling and the final sample consisted of N=7. The sample consisted of children with intellectual disability between the age span of six and nine of both male and female gender. The experiment consisted of three parts: a pretest where data in terms of the socio-emotional skill of emotion recognition was collected, the actual intervention where the participants engaged inplaying a digital game, and lastly, a posttest where the skill of emotion recognition was measured again. The data was subsequently analyzed via a paired sample t-test. The results of the study showed that DGBL did not significantly increase the socio-emotional skill of emotion recognition in children with intellectual disability. This result in part contradicts earlier research on DGBL and intellectual disability as well as DGBL andsocio-emotional skill where significant effects have been identified. However, since no previous research has investigated whether DGBL could be effective in increasing the socio-emotional skill of children with intellectual disability, future research is needed to confirm or reject the present results. In summary, the current research has extended on current knowledge and provided important implications for the field of special education.