Dans cette thèse, nous explorons les concepts de l'intelligence ambiante et les mécanismes facilitant leurs usages pour la conception d'environnements de vente intelligents. Une partie du travail fût réalisée dans le cadre du projet ANR p-LearNet.Notre objectif est de démontrer la faisabilité d'une solution permettant la génération d'espaces de ventes intelligents.Nous avons alors réalisé une usine logicielle. Cette usine a pour objectif, à partir de modèles, de générer tout ou partie de code permettant l'exécution de l'espace de vente. Ce code peut être exécuté sur une plate-forme de communication intégrée à l'usine.Nous avons dans un premier temps défini une infrastructure de communication permettant de gérer un ensemble de dispositifs hétérogènes. Cette plate-forme est un ESB léger dédié aux espaces de ventes intelligents. Cette plate-forme a été testée au cours du projet p-LearNet. Dans un second temps, nous avons créé des modèles qui représentent notre domaine d'application, les espaces de vente intelligents. Ces modèles permettent de définir les principaux éléments qui vont constituer un espace intelligent : les acteurs, les meubles, les services proposés. A partir de ces modèles, nous avons implémenté un éditeur graphique. Cet éditeur est à destination d'un utilisateur non-informaticien pour qu'il puisse graphiquement construire un espace de vente intelligent. Cet outil génère ensuite du code qui peut être exécuté sur la plate-forme décrite précédemment. / In this thesis, we explore the concepts of ambient intelligence and mechanisms to facilitate their use for designing intelligent selling environments. Part of the work was carried out within the ANR p-LearNet project.Our goal is to demonstrate the feasibility of a solution to the generation of intelligent selling spaces.Then We performed a software factory. This software factory aims to generate all or part of code to run the selling space from models. This code can be run on a communication platform integrated within the factory.We initially defined a communication infrastructure to manage a set of heterogeneous devices. This platform, based on OSGi and UPnP, is considered as a lightweight ESB dedicated to intelligent selling spaces. This platform has been tested during p-LearNet project and deployed within supermarkets. In a second step, we have created models that represent our scope, intelligent selling apces. These models define the key elements that will form a smart space: actors, furniture, services offered,…From these models, we have implemented a graphical editor. This editor is intended for a user, which knowledge in computer Sciences is poor, to graphically build a intelligent selling space. Then the tool generates code that can be executed on the platform described above. The use of this tool has been demonstrated through the implementation of a improved display casewindow. This showcase, presented at the salon VAD Lille, is running on the communication platform defined in the software factory.

Με τη συνεχώς αυξανόμενη παρουσία αυτόματων συστημάτων στην καθημερινότητά μας, εισέρχεται και το βάρος της αλληλεπίδρασης με αυτά τα συστήματα, εξαιτίας της έλλειψης συναισθηματικής νοημοσύνης από την πλευρά των μηχανών. Η συναισθηματική πληροφορία που μεταδίδεται μέσω της ανθρώπινης ομιλίας αποτελεί σημαντικό παράγοντα στην ανθρώπινη επικοινωνία και αλληλεπίδραση. Όταν όμως μηχανές ή υπολογιστικά συστήματα επικοινωνούν με ανθρώπους δεν αντιλαμβάνονται τις πιθανές συναισθηματικές καταστάσεις του ομιλητή και γι’ αυτό το λόγο υστερούν. Στα πλαίσια αυτής της εργασίας πραγματοποιείται επισημείωση, σε επίπεδο προτάσεων, λέξεων και ήχων, σε ηχογραφήσεις που έχουν πραγματοποιηθεί κατά την διάρκεια αλληλεπίδρασης ανθρώπου-μηχανής, με βάση την προβολή των συναισθημάτων στον κύκλο «ενεργοποίησης-αξιολόγησης». Έτσι καταλήγουμε στη δημιουργία μια ελληνικής βάσης συναισθηματικής ομιλίας. Στη συνέχεια, με βάση τις επισημειωμένες φράσεις εκπαιδεύουμε το γλωσσολογικό μας μοντέλο, το οποίο και ελέγχουμε με ένα υποσύνολο αυτών, για να δούμε κατά πόσο τα αποτελέσματα μας ανταποκρίνονται στην πραγματικότητα. Αποτέλεσμα της διαδικασίας είναι μια αριθμητική τιμή πρόβλεψης για την συναισθηματική κατάσταση της ομιλούμενης φράσης. / -

La dynamique du geste reste peu exploitée dans le domaine de l’interaction hommemachinealors qu’elle permet de définir des critères simples d’analyse de l’intention desutilisateurs. Sa principale utilisation porte sur les fonctions de transfert des périphériquesisotoniques. Ces fonctions permettent d’établir une relation entre le geste de l’utilisateureffectué dans l’espace moteur avec le périphérique de pointage (i.e. souris) et le mouvementdu pointeur dans l’espace visuel. Les fonctions de transfert utilisées par les systèmesd’exploitation modernes restent cependant fortement méconnues même si des études ontmontré qu’elles ont un impact sur les performances des utilisateurs. Pour caractériser cesfonctions, nous avons développé un périphérique électronique qui simule une souris réelleet qui permet d’obtenir les fonctions de transfert de tout système. Nous avons ainsi puobtenir les fonctions des systèmes Windows, Linux et Mac et nous avons observé des différencesnotables entre ces fonctions. Nous avons ensuite cherché à optimiser ces fonctionsen réduisant au maximum le nombre de leurs degrés de liberté. En faisant varier de manièreexhaustive ces paramètres, nous avons ainsi pu observé des différences significativesqui montrent qu’il existe une famille de fonctions pour lesquelles les performances sontoptimales. Enfin, nous nous sommes intéressés à l’utilisation de la dynamique du gestedans un autre contexte, celui de la génération d’événements discrets, pertinent dans lecadre des interfaces sans contact de type Kinect. Nous avons ainsi montré qu’il est possiblede distinguer la sélection du déplacement dans une tâche d’interaction 3D. La techniquedéveloppée est ensuite améliorée dans le cas des cibles de petites tailles et son applicationdans un contexte d’interaction 2D est évaluée. / The dynamic of the gesture is not used a lot in the domain of human-computer interaction,while it can help to define some simple heuristics to analyze the intents of users.Its first use is related to transfer functions for isotonic devices. Those functions establisha relationship between the gesture of the user, executed in motor space with the pointingdevice (i.e. a mouse), and the movement of the pointer in the visual space. The transferfunctions used by modern operating systems are still not well known, even if somestudies have shown they have an impact on user’s performance. To characterize thosefunctions, we have developed an electronic device which simulates a real mouse, and thatcan retrieve the transfer functions of any system. We have then exposed the functions ofthe Windows, Linux and Mac systems, and we observed some visible differences betweenthose functions. We have then tried to optimize those functions by reducing their numberof degrees of freedom. By varying those parameters in an exhaustive way, we have thenobserved some significant differences that show there is a family of functions for whichperformances are optimal. Finally, we were interested in the use of the dynamic of thegesture in another context : the generation of discrete events, useful when dealing withcontact-less interfaces, like the Kinect. We have shown that it is possible to distinguishbetween picking and drag’n’drop in a 3D interaction task. The technique is then improvedin the case of small size targets, and its application in a 2D interaction context isevaluated.

This thesis is concerned with understanding and describing the nature of 'community' within digital domains. A literature review indicates multiple media use within communities. The increasing range of personal and organisational technologies available suggests digital communities are more than just online communities. As such they require a new method of assessment. The design of digital communities should be based on an understanding of 'community' in digital domains. Previous assessments, often focusing exclusively on the Internet, failed to recognise the ways in which technologies are integrated within communities. A new assessment method should allow the examination of integrated technology effects on communities through an analysis of important community features. To assess digital communities a framework consisting of five headings was developed. The framework allows the effects of technologies to be examined across a range of communities. Taking a convergent methodologies approach five studies were undertaken covering a range of technologies and media integration issues. The results suggest that digital communities are groups of people using technology to support their social interaction needs. Media use within digital communities is heavily integrated and the social needs of community members drive technology use. Designers should provide communities with flexible technology that permits integration and member adaptation.

Cette thèse s'intéresse à comment utiliser les gestes dans l'air pour enrichir les l'interaction Homme-Machine en utilisant des espaces interactifs. Cette thèse s'inspire d'un concept de la communication non verbale : la proxémie. Cette théorie, introduite par Edward T. Hall, affirme entre autres que notre perception de l'espace est dynamique. Et s'accorde à l'environnement que nous percevons. En m'inspirant de cette théorie, je présente ici un nouveau concept d'interaction accompagné de son framework de design : Mimetic Interaction Space (MIS). Afin de montrer ce que le concept peut apporter à l'interaction, en plus d'une relecture de la littérature sur l'interaction dans les airs, je propose trois instanciations de ce concept autour des trois types d'utilisation du concept. La première instanciation est pour le contrôle indirect sur un écran distant en utilisant un MIS comme une interface à part entière. La seconde instanciation d'enrichir l'interaction sur tablette en utilisant un ou des MISs l'entourant. Deux propositions d'utilisation sont faites. Une première en subdivisant le MIS en plusieurs autour de la tablette. Puis une seconde utilisation du MIS comme étant la continuité de l'écran de la tablette. La troisième instanciation se fait dans le contexte de l'interaction sur très grands écrans tactiles. Ici, un MIS a pour rôle de faire la transition continue entre l'interaction tactile et l'interaction dans les airs. Pour finir, j'introduis quelques pistes de développement pour l'avenir des MIS et je propose une réflexion sur une facette du concept des MIS qui ouvre d'importantes questions sur l'interaction basée MIS. / This dissertation focuses on how to exploit gestural midair interaction to extend the possibilities of existing devices by using interactive spaces. The starting point is in the nonverbal communication theory of proxemics introduced by Eward T. Hall who stated that our perception of space is dynamic. From this, I argue that we could apply this dynamic understanding of space to interactive spaces. I propose a novel concept of interaction and an associated design framework for interactive spaces : Mimetic Interaction Space (MIS). To show the prospects MIS gives for midair interaction, I propose three instantiations of the concept that uses it in different ways. The first one is the use of MISs as a standalone interface the control of a remote display. The second instantiation is the use of one or several MIS tied up to the tablet in two ways. First by cutting out the MIS in multiple ones. The second way of using a MIS linked to the tablet is by considering it as a continuation of the tablet screen around it.The third instantiation is in the context of interaction on wall displays where a MIS is placed right in front of the screen and has the role of a transition space from touch to midair interaction. This MIS allows for a continuous transition between the physical and direct nature of touch interaction, and the more abstract nature of midair interaction. I finally conclude by discussing the future of interfaces regarding midair gestures. I also discuss a facet of MIS that opens a novel way to think about MIS interaction.

The research in this thesis was motivated by an interest in understanding the work and effort that goes into organising family homes, with the aim of informing the design of novel information technology for the home. It was undertaken to address a notable absence of in-depth research into domestic information and communication technology in the fields of Human Computer Interaction (HCI) and Computer Supported Cooperative Work (CSCW). To that end, this thesis presents an ethnographic study of everyday routines in thirteen family homes. Following an established tradition within HCI and CSCW, the study applies qualitative fieldwork methods as a means to investigate and interpret the empirical materials. Periods of extended observation and semi-structured interviews with the thirteen families over a three-year period form the basis of the empirical material. The materials are analysed using a hybrid perspective composed of a combination of influences from the study of material culture, to interaction analysis and ethnography. The hybrid analytical perspective draws out insights regarding the families’ mundane practices and the artfully devised solutions they use to organise daily life. Four household activities and artefacts are given specific focus: (i) household list making, (ii) the display qualities of refrigerator doors, (iii) the organisation of household clutter, and (iv) the devising of bespoke solutions in organising home life. Broader findings include the observations that people tailor solutions to meet their needs, that optimum efficiency is not the pre-eminent determinant in what method or artefact people choose to organise themselves and their homes, and that homes determine their individual characters in part by how everyday tasks and organisation are accomplished. In short, the personal qualities of these mundane practices are part of what makes a home a home. These findings are used to elicit implications for information technology design, with the aim of encouraging designers of domestic technology to be aware of and respectful towards the idiosyncratic nature of the home, and, wherever possible, to design in such a way as to allow the technology to be appropriated for families’ bespoke tailoring. To evaluate and address this point, two design projects, one on augmented magnets and another on a “media bowl”, are used to develop and test out this approach. Both projects are critically examined to reflect on the efficacy of the design approach and what lessons might be learnt for future studies and design exercises. The combination of detailed ethnographic fieldwork on family homes combined with the development of experimental design projects is intended to deepen the understanding of the mundane behaviours and everyday routines of family homes, in order to better inform the design of information technology for the home.

This work investigates impression management and in particular impression management using ubiquitous technology. Generally impression management is the process through which people try to influence the impressions that others have about them. In particular, impression management focuses on the flow of information between a performer and his/her audience, with control over what is presented to whom being of the utmost importance when trying to create the appropriate impression. Ubiquitous technology has provided opportunities for individuals to present themselves to others. However, the disconnection between presenter and audience over both time and space can result in individuals being misrepresented. This thesis outlines two important areas when trying to control the impression one gives namely, hiding and revealing, and accountability. By exploring these two themes the continuous evolution and dynamic nature of controlling the impression one gives is explored. While this ongoing adaptation is recognised by designers they do not always create technology that is sufficiently dynamic to support this process. As a result, this work attempts to answer three research questions: RQ1: How do users of ubicomp systems appropriate recorded data from their everyday activity and make it into a resource for expressing themselves to others in ways that are dynamically tailored to their ongoing social context and audience? RQ2: What technology can be built to support ubicomp system developers to design and develop systems to support appropriation as a central part of a useful or enjoyable user experience? RQ3: What software architectures best suit this type of appropriated interaction and developers’ designing to support such interaction? Through a thorough review of existing literature, and the extensive study of several large ubicomp systems, the issues when presenting oneself through technology are identified. The main issues identified are hiding and revealing, and accountability. These are built into a framework that acts as a reference for designers wishing to support impression management. An architecture for supporting impression management has also been developed that conforms to this framework and its evolution is documented later in the thesis. A demonstration of this architecture in a multi-player mobile experience is subsequently presented.

- / -

'Complex sociotechnical systems' are systems made up of numerous interacting parts, both human and non-human, operating in dynamic, ambiguous and safety critical domains. The system design, and representation, has significant implications for the usability of the system, its performance, errors and reliability. Due to its formative constraint-based approach, Cognitive Work Analysis (CWA) has been frequently proposed as a framework for modelling these complex systems, further it is contended that the approach can be extended to design for interaction without significant deviation from the accepted framework. This thesis presents a number of applications in complex military domains to explore and develop the benefits of CWA. Unlike much of the previous literature, particular attention is placed on exploring the CWA framework in its entirety. This holistic approach focuses on the system environment, the activity that takes place within it, the strategies used to conduct this activity, the way in which the constituent parts of the system (both human and non-human) interact and the behaviour required. Each stage of this analysis identifies the constraints governing the system; it is contended that through this holistic understanding of constraints, recommendations can be made for the design of system interaction; increasing the ability of users to cope with unanticipated, unexpected situations. This thesis discusses the applicability of the approach in system analysis, development and evaluation. It provides process to what was previously a loosely defined framework.

In this thesis, we describe extensive experiments on the classification of emotions from speech using acoustic features. This area of research has important applications in human computer interaction. We have thoroughly reviewed the current literature and present our results on some of the contemporary emotional speech databases. The principal focus is on creating a large set of acoustic features, descriptive of different emotional states and finding methods for selecting a subset of best performing features by using feature selection methods. In this thesis we have looked at several traditional feature selection methods and propose a novel scheme which employs a preferential Borda voting strategy for ranking features. The comparative results show that our proposed scheme can strike a balance between accurate but computationally intensive wrapper methods and less accurate but computationally less intensive filter methods for feature selection. By using the selected features, several schemes for extending the binary classifiers to multiclass classification are tested. Some of these classifiers form serial combinations of binary classifiers while others use a hierarchical structure to perform this task. We describe a new hierarchical classification scheme, which we call Data-Driven Dimensional Emotion Classification (3DEC), whose decision hierarchy is based on non-metric multidimensional scaling (NMDS) of the data. This method of creating a hierarchical structure for the classification of emotion classes gives significant improvements over other methods tested. The NMDS representation of emotional speech data can be interpreted in terms of the well-known valence-arousal model of emotion. We find that this model does not give a particularly good fit to the data: although the arousal dimension can be identified easily, valence is not well represented in the transformed data. From the recognition results on these two dimensions, we conclude that valence and arousal dimensions are not orthogonal to each other. In the last part of this thesis, we deal with the very difficult but important topic of improving the generalisation capabilities of speech emotion recognition (SER) systems over different speakers and recording environments. This topic has been generally overlooked in the current research in this area. First we try the traditional methods used in automatic speech recognition (ASR) systems for improving the generalisation of SER in intra– and inter–database emotion classification. These traditional methods do improve the average accuracy of the emotion classifier. In this thesis, we identify these differences in the training and test data, due to speakers and acoustic environments, as a covariate shift. This shift is minimised by using importance weighting algorithms from the emerging field of transfer learning to guide the learning algorithm towards that training data which gives better representation of testing data. Our results show that importance weighting algorithms can be used to minimise the differences between the training and testing data. We also test the effectiveness of importance weighting algorithms on inter–database and cross-lingual emotion recognition. From these results, we draw conclusions about the universal nature of emotions across different languages.