This master’s thesis is dealing with recognition of an easy static gestures in order to computer controlling. First part of this work is attended to the theoretical review of methods used to hand segmentation from the image. Next methods for hang gesture classification are described. The second part of this work is devoted to choice of suitable method for hand segmentation based on skin color and movement. Methods for hand gesture classification are described in next part. Last part of this work is devoted to description of proposed system.

Tesis por compendio / [EN] The main objective of this thesis is the development of new sensors and actuators using Printed Electronics technology. For this, conductive, semiconductor and dielectric polymeric materials are used on flexible and/or elastic substrates. By means of suitable designs and application processes, it is possible to manufacture sensors capable of interacting with the environment. In this way, specific sensing functionalities can be incorporated into the substrates, such as textile fabrics. Additionally, it is necessary to include electronic systems capable of processing the data obtained, as well as its registration. In the development of these sensors and actuators, the physical properties of the different materials are precisely combined. For this, multilayer structures are designed where the properties of some materials interact with those of others. The result is a sensor capable of capturing physical variations of the environment, and convert them into signals that can be processed, and finally transformed into data. On the one hand, a tactile sensor printed on textile substrate for 2D gesture recognition was developed. This sensor consists of a matrix composed of small capacitive sensors based on a capacitor type structure. These sensors were designed in such a way that, if a finger or other object with capacitive properties, gets close enough, its behaviour varies, and it can be measured. The small sensors are arranged in this matrix as in a grid. Each sensor has a position that is determined by a row and a column. The capacity of each small sensor is periodically measured in order to assess whether significant variations have been produced. For this, it is necessary to convert the sensor capacity into a value that is subsequently digitally processed. On the other hand, to improve the effectiveness in the use of the developed 2D touch sensors, the way of incorporating an actuator system was studied. Thereby, the user receives feedback that the order or action was recognized. To achieve this, the capacitive sensor grid was complemented with an electroluminescent screen printed as well. The final prototype offers a solution that combines a 2D tactile sensor with an electroluminescent actuator on a printed textile substrate. Next, the development of a 3D gesture sensor was carried out using a combination of sensors also printed on textile substrate. In this type of 3D sensor, a signal is sent generating an electric field on the sensors. This is done using a transmission electrode located very close to them. The generated field is received by the reception sensors and converted to electrical signals. For this, the sensors are based on electrodes that act as receivers. If a person places their hands within the emission area, a disturbance of the electric field lines is created. This is due to the deviation of the lines to ground using the intrinsic conductivity of the human body. This disturbance affects the signals received by the electrodes. Variations captured by all electrodes are processed together and can determine the position and movement of the hand on the sensor surface. Finally, the development of an improved 3D gesture sensor was carried out. As in the previous development, the sensor allows contactless gesture detection, but increasing the detection range. In addition to printed electronic technology, two other textile manufacturing technologies were evaluated. / [ES] La presente tesis doctoral tiene como objetivo fundamental el desarrollo de nuevos sensores y actuadores empleando la tecnología electrónica impresa, también conocida como Printed Electronics. Para ello, se emplean materiales poliméricos conductores, semiconductores y dieléctricos sobre sustratos flexibles y/o elásticos. Por medio de diseños y procesos de aplicación adecuados, es posible fabricar sensores capaces de interactuar con el entorno. De este modo, se pueden incorporar a los sustratos, como puedan ser tejidos textiles, funcionalidades específicas de medición del entorno y de respuesta ante cambios de este. Adicionalmente, es necesario incluir sistemas electrónicos, capaces de realizar el procesado de los datos obtenidos, así como de su registro. En el desarrollo de estos sensores y actuadores se combinan las propiedades físicas de los diferentes materiales de forma precisa. Para ello, se diseñan estructuras multicapa donde las propiedades de unos materiales interaccionan con las de los demás. El resultado es un sensor capaz de captar variaciones físicas del entorno, y convertirlas en señales que pueden ser procesadas y transformadas finalmente en datos. Por una parte, se ha desarrollado un sensor táctil impreso sobre sustrato textil para reconocimiento de gestos en 2D. Este sensor se compone de una matriz formada por pequeños sensores capacitivos basados en estructura de tipo condensador. Estos se han diseñado de forma que, si un dedo u otro objeto con propiedades capacitivas se aproxima suficientemente, su comportamiento varía, pudiendo ser medido. Los pequeños sensores están ordenados en dicha matriz como en una cuadrícula. Cada sensor tiene una posición que viene determinada por una fila y por una columna. Periódicamente se mide la capacidad de cada pequeño sensor con el fin de evaluar si ha sufrido variaciones significativas. Para ello es necesario convertir la capacidad del sensor en un valor que posteriormente es procesado digitalmente. Por otro lado, con el fin de mejorar la efectividad en el uso de los sensores táctiles 2D desarrollados, se ha estudiado el modo de incorporar un sistema actuador. De esta forma, el usuario recibe una retroalimentación indicando que la orden o acción ha sido reconocida. Para ello, se ha complementado la matriz de sensores capacitivos con una pantalla electroluminiscente también impresa. El resultado final ofrece una solución que combina un sensor táctil 2D con un actuador electroluminiscente realizado mediante impresión electrónica sobre sustrato textil. Posteriormente, se ha llevado a cabo el desarrollo de un sensor de gestos 3D empleando una combinación de sensores impresos también sobre sustrato textil. En este tipo de sensor 3D, se envía una señal que genera un campo eléctrico sobre los sensores impresos. Esto se lleva a cabo mediante un electrodo de transmisión situado muy cerca de ellos. El campo generado es recibido por los sensores y convertido a señales eléctricas. Para ello, los sensores se basan en electrodos que actúan de receptores. Si una persona coloca su mano dentro del área de emisión, se crea una perturbación de las líneas de los campos eléctricos. Esto es debido a la desviación de las líneas de campo a tierra utilizando la conductividad intrínseca del cuerpo humano. Esta perturbación cambia/afecta a las señales recibidas por los electrodos. Las variaciones captadas por todos los electrodos son procesadas de forma conjunta pudiendo determinar la posición y el movimiento de la mano sobre la superficie del sensor. Finalmente, se ha llevado a cabo el desarrollo de un sensor de gestos 3D mejorado. Al igual que el desarrollo anterior, permite la detección de gestos sin necesidad de contacto, pero incrementando la distancia de alcance. Además de la tecnología de impresión electrónica, se ha evaluado el empleo de otras dos tecnologías de fabricación textil. / [CA] La present tesi doctoral té com a objectiu fonamental el desenvolupament de nous sensors i actuadors fent servir la tecnologia de electrònica impresa, també coneguda com Printed Electronics. Es va fer us de materials polimèrics conductors, semiconductors i dielèctrics sobre substrats flexibles i/o elàstics. Per mitjà de dissenys i processos d'aplicació adequats, és possible fabricar sensors capaços d'interactuar amb l'entorn. D'aquesta manera, es poden incorporar als substrats, com ara teixits tèxtils, funcionalitats específiques de mesurament de l'entorn i de resposta davant canvis d'aquest. Addicionalment, és necessari incloure sistemes electrònics, capaços de realitzar el processament de les dades obtingudes, així com del seu registre. En el desenvolupament d'aquests sensors i actuadors es combinen les propietats físiques dels diferents materials de forma precisa. Cal dissenyar estructures multicapa on les propietats d'uns materials interaccionen amb les de la resta. manera El resultat es un sensor capaç de captar variacions físiques de l'entorn, i convertirles en senyals que poden ser processades i convertides en dades. D'una banda, s'ha desenvolupat un sensor tàctil imprès sobre substrat tèxtil per a reconeixement de gestos en 2D. Aquest sensor es compon d'una matriu formada amb petits sensors capacitius basats en una estructura de tipus condensador. Aquests s'han dissenyat de manera que, si un dit o un altre objecte amb propietats capacitives s'aproxima prou, el seu comportament varia, podent ser mesurat. Els petits sensors estan ordenats en aquesta matriu com en una quadrícula. Cada sensor té una posició que ve determinada per una fila i per una columna. Periòdicament es mesura la capacitat de cada petit sensor per tal d'avaluar si ha sofert variacions significatives. Per a això cal convertir la capacitat del sensor a un valor que posteriorment és processat digitalment. D'altra banda, per tal de millorar l'efectivitat en l'ús dels sensors tàctils 2D desenvolupats, s'ha estudiat la manera d'incorporar un sistema actuador. D'aquesta forma, l'usuari rep una retroalimentació indicant que l'ordre o acció ha estat reconeguda. Per a això, s'ha complementat la matriu de sensors capacitius amb una pantalla electroluminescent també impresa. El resultat final ofereix una solució que combina un sensor tàctil 2D amb un actuador electroluminescent realitzat mitjançant impressió electrònica sobre substrat tèxtil. Posteriorment, s'ha dut a terme el desenvolupament d'un sensor de gestos 3D emprant una combinació d'un mínim de sensors impresos també sobre substrat tèxtil. En aquest tipus de sensor 3D, s'envia un senyal que genera un camp elèctric sobre els sensors impresos. Això es porta a terme mitjançant un elèctrode de transmissió situat molt a proper a ells. El camp generat és rebut pels sensors i convertit a senyals elèctrics. Per això, els sensors es basen en elèctrodes que actuen de receptors. Si una persona col·loca la seva mà dins de l'àrea d'emissió, es crea una pertorbació de les línies dels camps elèctrics. Això és a causa de la desviació de les línies de camp a terra utilitzant la conductivitat intrínseca de el cos humà. Aquesta pertorbació afecta als senyals rebudes pels elèctrodes. Les variacions captades per tots els elèctrodes són processades de manera conjunta per determinar la posició i el moviment de la mà sobre la superfície del sensor. Finalment, s'ha dut a terme el desenvolupament d'un sensor de gestos 3D millorat. A l'igual que el desenvolupament anterior, permet la detecció de gestos sense necessitat de contacte, però incrementant la distància d'abast. A més a més de la tecnologia d'impressió electrònica, s'ha avaluat emprar altres dues tecnologies de fabricació tèxtil. / Ferri Pascual, J. (2020). Tactile and Touchless Sensors Printed on Flexible Textile Substrates for Gesture Recognition [Tesis doctoral]. Universitat Politècnica de València. https://doi.org/10.4995/Thesis/10251/153075 / Compendio

This master thesis presents a machine learning framework for real-time gesture and skeleton-based action recognition, integrated with the Unity game engine. The system aims to enhance human-computer interaction (HCI) in gaming and 3D related applications through natural movement recognition, by training a model on skeleton tracking data. The framework is trained to accurately categorize and identify gestures such as kicks and punches, enabling a more immersive gaming experience not existing in traditional controllers. After studying the evolution of HCI and how machine learning has transformed and reshaped the interaction paradigm, the prototype system is built through data collection, augmenting, and preprocessing, followed by training and evaluating a Long Short-Term Memory (LSTM) neural network model for gesture classification. The model is integrated into Unity via Unity Sentis using Open Neural Network Exchange (ONNX) format, enabling efficient real-time action recognition in 3D space. Each component of the pipeline is available and adaptable for future custom- ization and needs, skeleton tracking and Unity integration is built using the ZED 2i camera and ZED SDK. Experimental results demonstrate that the system presented can achieve over 90% accuracy in identifying predefined gestures. As a bridging solution tailored for Unity, this framework offers a practical solution to action recognition that could be found useful in future applications. This work contributes to advancing human-computer interaction and offers a foundation for further development in gesture-based Unity game design.

La présente thèse a pour but d’approfondir l’étude du geste dans le cadre de l’interaction Homme Machine. Il s’agit de créer de nouveaux paradigmes d’interaction qui offrent à l’utilisateur de plus amples possibilités d’expression basées sur le geste. Un des vecteurs d’expression du geste, très rarement traité en Interaction Homme Machine, qui lui confère sa coloration et son aspect, est ce que les théoriciens et praticiens de la danse appellent « les qualités de mouvement ». Nous mettons à profit des collaborations avec le domaine de la danse pour étudier la notion de qualités de mouvement et l’intégrer à des paradigmes d’interaction gestuelle. Notre travail analyse les apports de l’intégration des qualités de mouvement comme modalité d’interaction, fournit les outils propices à l’élaboration de cette intégration (en termes de méthodes d’analyse, de visualisation et de contrôle gestuel), en développe et évalue certaines techniques d’interaction.Les contributions de la thèse se situent d’abord dans la formalisation de la notion de qualités de mouvement et l’évaluation de son intégration dans un dispositif interactif en termes d’expérience utilisateur. Sur le plan de la visualisation des qualités de mouvement, les travaux menés pendant la thèse ont permis de démontrer que les modèles physiques masses-ressorts offrent de grandes possibilités de simulation de comportements dynamiques et de contrôle en temps réel. Sur le plan de l’analyse, la thèse a permis de développer des approches novatrices de reconnaissance automatique des qualités de mouvement de l’utilisateur. Enfin, à partir des approches d’analyse et de visualisation des qualités de mouvement, la thèse a donné lieu à l’implémentation d’un ensemble de techniques d’interaction. Elle a appliqué et évalué ses techniques dans le contexte de la pédagogie de la danse et de la performance. / The thesis studies gesture in the context of Human-Computer interaction. It aims at creating new interaction paradigms that offer the user further expressive possibilities based on gestures. The theorists and practitioners of the dance call "movement qualities” (MQ), a notion that conveys expressive content describing the way a gesture is performed. This notion has been rarely taken into consideration in the field of HCI. Our work draws on collaborations with the field of dance to explore the notion of movement qualities and to integrate it as interaction modality. 

The contributions of the thesis are in the formalism of the notion of movement qualities and evaluation of its integration as interaction modality in terms of user experience. 

We also provide computational tools for considering MQ in interactive systems in terms of analysis, representation and gesture control methods. On the representational level, our work have demonstrated that physical models based on masses and springs systems offer great opportunities for simulating dynamics related to MQs and for real-time gesture control. On the analysis level, we developed innovative approaches to automatic real time recognition of movement qualities. Finally, we implemented of a set of interaction techniques based on movement qualities that we applied and evaluated in the context of dance pedagogy and performance.

Cette thèse explore la reconnaissance de gestes à partir de capteurs inertiels pour Smartphone. Ces gestes consistent en la réalisation d'un tracé dans l'espace présentant une valeur sémantique, avec l'appareil en main. Notre étude porte en particulier sur l'apprentissage de métrique entre signatures gestuelles grâce à l'architecture "Siamoise" (réseau de neurones siamois, SNN), qui a pour but de modéliser les relations sémantiques entre classes afin d'extraire des caractéristiques discriminantes. Cette architecture est appliquée au perceptron multicouche (MultiLayer Perceptron). Les stratégies classiques de formation d'ensembles d'apprentissage sont essentiellement basées sur des paires similaires et dissimilaires, ou des triplets formés d'une référence et de deux échantillons respectivement similaires et dissimilaires à cette référence. Ainsi, nous proposons une généralisation de ces approches dans un cadre de classification, où chaque ensemble d'apprentissage est composé d’une référence, un exemple positif, et un exemple négatif pour chaque classe dissimilaire. Par ailleurs, nous appliquons une régularisation sur les sorties du réseau au cours de l'apprentissage afin de limiter les variations de la norme moyenne des vecteurs caractéristiques obtenus. Enfin, nous proposons une redéfinition du problème angulaire par une adaptation de la notion de « sinus polaire », aboutissant à une analyse en composantes indépendantes non-linéaire supervisée. A l'aide de deux bases de données inertielles, la base MHAD (Multimodal Human Activity Dataset) ainsi que la base Orange, composée de gestes symboliques inertiels réalisés avec un Smartphone, les performances de chaque contribution sont caractérisées. Ainsi, des protocoles modélisant un monde ouvert, qui comprend des gestes inconnus par le système, mettent en évidence les meilleures capacités de détection et rejet de nouveauté du SNN. En résumé, le SNN proposé permet de réaliser un apprentissage supervisé de métrique de similarité non-linéaire, qui extrait des vecteurs caractéristiques discriminants, améliorant conjointement la classification et le rejet de gestes inertiels. / As consumer devices become more and more ubiquitous, new interaction solutions are required. In this thesis, we explore inertial-based gesture recognition on Smartphones, where gestures holding a semantic value are drawn in the air with the device in hand. In our research, speed and delay constraints required by an application are critical, leading us to the choice of neural-based models. Thus, our work focuses on metric learning between gesture sample signatures using the "Siamese" architecture (Siamese Neural Network, SNN), which aims at modelling semantic relations between classes to extract discriminative features, applied to the MultiLayer Perceptron. Contrary to some popular versions of this algorithm, we opt for a strategy that does not require additional parameter fine tuning, namely a set threshold on dissimilar outputs, during training. Indeed, after a preprocessing step where the data is filtered and normalised spatially and temporally, the SNN is trained from sets of samples, composed of similar and dissimilar examples, to compute a higher-level representation of the gesture, where features are collinear for similar gestures, and orthogonal for dissimilar ones. While the original model already works for classification, multiple mathematical problems which can impair its learning capabilities are identified. Consequently, as opposed to the classical similar or dissimilar pair; or reference, similar and dissimilar sample triplet input set selection strategies, we propose to include samples from every available dissimilar classes, resulting in a better structuring of the output space. Moreover, we apply a regularisation on the outputs to better determine the objective function. Furthermore, the notion of polar sine enables a redefinition of the angular problem by maximising a normalised volume induced by the outputs of the reference and dissimilar samples, which effectively results in a Supervised Non-Linear Independent Component Analysis. Finally, we assess the unexplored potential of the Siamese network and its higher-level representation for novelty and error detection and rejection. With the help of two real-world inertial datasets, the Multimodal Human Activity Dataset as well as the Orange Dataset, specifically gathered for the Smartphone inertial symbolic gesture interaction paradigm, we characterise the performance of each contribution, and prove the higher novelty detection and rejection rate of our model, with protocols aiming at modelling unknown gestures and open world configurations. To summarise, the proposed SNN allows for supervised non-linear similarity metric learning, which extracts discriminative features, improving both inertial gesture classification and rejection.

Made available in DSpace on 2016-12-12T20:22:52Z (GMT). No. of bitstreams: 1 RAFAEL ALCESTE BERRI.pdf: 28428368 bytes, checksum: 667b9facc9809bfd5e0847e15279b0e6 (MD5) Previous issue date: 2014-02-21 / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior / In this work, three proposals of systems have been developed using a frontal camera to monitor the driver and enabling to identificate if a cell phone is being used while driving the vehicle. It is estimated that 80% of crashes and 65% of near collisions involved drivers who were inattentive in traffic for three seconds before the event. Five videos in real environment were generated to test the systems. The pattern recognition system (RP) uses adaptive skin segmentation, feature extraction, and machine learning to detect cell phone usage on each frame. The cell phone detection happens when, in periods of 3 seconds, 60% (threshold) of frames or more are identified as a cell phone use, individually. The average accuracy on videos achieved was 87.25% with Multilayer Perceptron (MLP), Gaussian activation function, and two neurons of the intermediate layer. The movement detection system (DM) uses optical flow, filtering the most relevant movements of the scene, and three successive frames for detecting the movements to take the phone to the ear and take it off. The DM proposal was not demonstrated as being an effective solution for detecting cell phone use, reaching an accuracy of 52.86%. The third solution is a hybrid system. It uses the RP system for classification and the DM for choosing the RP parameters. The parameters chosen for RP are the threshold and the classification system. The definition of these two parameters occurs at the end of each period, based on movement detected by the DM. Experimentally it was established that, when the movement induces to use cell phone, it is proper to use the threshold of 60%, and the classifier as MLP/Gaussian with seven neurons of the intermediate layer; otherwise, it is used threshold 85%, and MLP/Gaussian with two neurons of the intermediate layer for classification. The hybrid solution is the most robust system with average accuracy of 91.68% in real environment. / Neste trabalho, são desenvolvidas três propostas de sistemas que permitem identificar o uso de celular, durante o ato de dirigir um veículo, utilizando imagens capturadas de uma câmera posicionada em frente ao motorista. Estima-se que 80% das colisões e 65% das quase colisões envolveram motoristas que não estavam prestando a devida atenção ao trânsito por três segundos antes do evento. Cinco vídeos em ambiente real foram gerados com o intuito de testar os sistemas. A proposta de reconhecimento de padrões (RP) emprega segmentação de pele adaptativa, extração de características e aprendizado de máquina (classificador) na detecção do celular em cada quadro processado. A detecção do uso do celular ocorre quando, em períodos de 3 segundos, ao menos em 60% dos quadros (corte) são identificados com celular. A acurácia média nos vídeos alcançou 87, 25% ao utilizar Perceptron Multi-camadas (MLP) com função de ativação gaussiana e dois neurônios na camada intermediária como classificador. A proposta de detecção de movimento (DM) utiliza o fluxo ótico, filtragem dos movimentos mais relevantes da cena e três quadros consecutivos para detectar os momentos de levar o celular ao ouvido e o retirá-lo. A aplicação do DM, como solução para detectar o uso do celular, não se demostrou eficaz atingindo uma acurácia de 52, 86%. A terceira proposta, uma solução híbrida, utiliza o sistema RP como classificador e o de DM como seu parametrizador. Os parâmetros escolhidos para o sistema de RP são o corte e o sistema classificador. A definição desses dois parâmetros ocorre ao final de cada período, baseada na movimentação detectada pela DM. Com experimentações definiu-se que, caso a movimentação induza ao uso do celular, é adequado o uso do corte de 60% e o classificador MLP/Gaussiana com sete neurônios na camada intermediária, caso contrário, utiliza-se o corte de 85% e classificador MLP/Gaussiana com dois neurônios na mesma camada. A versão híbrida é a solução desenvolvida mais robusta, atingindo a melhor acurácia média de 91, 68% em ambiente real.

The dissemination of new innovative technology requires feasibility and simplicity. The problem with marker-based augmented reality is similar to glove-based hand gesture recognition: they both require an additional component to function. This thesis investigates the possibility of combining markerless augmented reality together with appearance-based hand gesture recognition by implementing a game with good user experience. The methods employed in this research consist of a game implementation and a pre-study meant for measuring interactive accuracy and precision, and for deciding upon which gestures should be utilized in the game. A test environment was realized in Unity using ARKit and Manomotion SDK. Similarly, the implementation of the game used the same development tools. However, Blender was used for creating the 3D models. The results from 15 testers showed that the pinching gesture was the most favorable one. The game was evaluated with a System Usability Scale (SUS) and received a score of 70.77 among 12 game testers, which indicates that the augmented reality game, which interaction method is solely based on bare-hands, can be quite enjoyable.