Les objets connectés ont aujourd’hui pénétré les foyers et, poussés par une société tournée de plus en plus vers le bien-être, ces capteurs mesurent et proposent dorénavant une grande variété de données physiologiques. L’arrivée à maturité des technologies de la réalité virtuelle, couplée avec l’avènement des objets connectés, permet et favorise dès lors de nouvelles perspectives dans la proposition d’expériences immersives enrichies. De nombreux travaux dans le domaine de la réalité virtuelle rapportent l’exploitation des signaux physiologiques. Ceux-ci se basent principalement sur du matériel médical, qui possède des contraintes d’utilisation forte, reste souvent encombrant et limite de fait la mobilité des utilisateurs. Pour tenter de pallier ces limites, nos travaux se concentrent donc sur l’utilisation originale des wearables (objets connectés portés) comme substituts aux capteurs physiologiques traditionnels dans le cadre d’applications immersives. Ce travail de thèse se positionne à mi-chemin entre une étude de faisabilité technologique et une étude fondamentale sur l’expérience utilisateur (UX).Dans ce contexte, l’objectif de nos recherches est de contribuer à la connaissance concernant l’impact de l’utilisation des données physiologiques dans des environnements virtuels immersifs. Nous étudierons en particulier l’influence d’un biofeedback cardiaque, via des capteurs connectés grand public, sur l’engagement utilisateur et le sentiment d’agentivité. Nous avons ainsi mené deux expérimentations nous permettant d’étudier l’impact des différentes modalités de biofeedback sur l’expérience utilisateur. Notre première expérimentation met en place un biofeedback cardiaque dans un jeu d’horreur en réalité virtuelle, permettant d’augmenter le sentiment de peur. Les résultats de cette expérimentationconfortent l’intérêt de l’utilisation de capteurs connectés comme moyen de captation physiologique dans des expériences de réalité virtuelle immersive. Ils mettent également en avant l’impact positif de ce biofeedback sur la dimension d’engagement de l’expérience utilisateur. La deuxième expérience porte sur l’utilisation de l’activité cardiaque comme une mécanique d’interaction obligatoire. Elle est découpée en deux parties, la première permettant de quantifier le niveau de compétence des participants dans le contrôle de leur activité cardiaque et la seconde les plongeant dans une suite de tâches en réalité virtuelle ; le contrôle cardiaque est de fait nécessaire pour les réussir. Les résultats de cette expérience démontrent la possibilité d’utiliser la dite mécanique pour des expériences virtuelles immersives et indiquent un impact positif sur le sentiment d’agentivité, lié au niveau de compétence des participants. Sur un plan théorique, cette thèse propose une synthèse des modèles de l’expérience utilisateur en environnement virtuel et soumet par ailleurs les bases d’un modèle que nous nommons « l’immersion physiologique ». / The internet of things has now entered every home and, with a society more and more focused towards wellness, these sensors measure and offer henceforth a wide variety of physiological data. Virtual reality technologies reaching maturity, coupled with the advent of the internet of things, allow consequently new opportunities to propose improved immersive experiences. If we identify nowadays many virtual reality studies reporting the usage of physiological data, they mainly use medical equipment, which poses strong usability constraints, is often cumbersome and limits mobility. In an attempt to overcome these limitations, this study therefore focuses on the original usage of smart wearables as substitutes for traditional sensors in immersive applications. Thus, this thesis is positioned halfway between a technological feasibility study and a fundamental user experience study.In this context, the objective of our study is to contribute to knowledge about the impact of the use of physiological data in immersive virtual environments. More precisely, the impact of biofeedback, via off-the-shelf smart wearables, on user engagement and the sense of agency. We have thus carried out two experiments allowing us to study the impacts of the different biofeedback modalities on user experience. Our first experiment implements a biofeedback based on heart rate in a virtual reality horror game, allowing to enhance the feeling of fear. The results of this experiment confirm the interest of using smart wearables to capture physiological data for immersive virtual reality experiences. They also highlight the positive impact of this biofeedback on user engagement. The second experiment focuses on the use of cardiac activity as a mandatory interaction mechanism. This experiment is divided into two parts, the first one quantifying the participants’ level of competency in heartrate control and the second one immersing them in a series of tasks in virtual reality ; heartrate control is necessary to complete the different steps of the experience. The results of this experiment demonstrate the possibility of using the said interaction mechanic for virtual reality experiences and indicate a positive impact on the sense of agency, linked with the level of competency of the participants. On a theoretical level, this thesis proposes a synthesis of user experience models in virtual environment and submit the foundations of a model that we call "physiological immersion".

Indiana University-Purdue University Indianapolis (IUPUI) / Smart wearables, known as emerging paradigms for vital big data capturing, have been attracting intensive attentions. However, one crucial problem is their power-hungriness, i.e., the continuous data streaming consumes energy dramatically and requires devices to be frequently charged. Targeting this obstacle, we propose to investigate the biodynamic patterns in the data and design a data-driven approach for intelligent data compression. We leverage Deep Learning (DL), more specifically, Convolutional Autoencoder (CAE), to learn a sparse representation of the vital big data. The minimized energy need, even taking into consideration the CAE-induced overhead, is tremendously lower than the original energy need. Further, compared with state-of-the-art wavelet compression-based method, our method can compress the data with a dramatically lower error for a similar energy budget. Our experiments and the validated approach are expected to boost the energy efficiency of wearables, and thus greatly advance ubiquitous big data applications in era of smart health. In recent years, there has also been a growing interest in edge intelligence for emerging instantaneous big data inference. However, the inference algorithms, especially deep learning, usually require heavy computation requirements, thereby greatly limiting their deployment on the edge. We take special interest in the smart health wearable big data mining and inference. Targeting the deep learning’s high computational complexity and large memory and energy requirements, new approaches are urged to make the deep learning algorithms ultra-efficient for wearable big data analysis. We propose to leverage knowledge distillation to achieve an ultra-efficient edge-deployable deep learning model. More specifically, through transferring the knowledge from a teacher model to the on-edge student model, the soft target distribution of the teacher model can be effectively learned by the student model. Besides, we propose to further introduce adversarial robustness to the student model, by stimulating the student model to correctly identify inputs that have adversarial perturbation. Experiments demonstrate that the knowledge distillation student model has comparable performance to the heavy teacher model but owns a substantially smaller model size. With adversarial learning, the student model has effectively preserved its robustness. In such a way, we have demonstrated the framework with knowledge distillation and adversarial learning can, not only advance ultra-efficient edge inference, but also preserve the robustness facing the perturbed input. / 2023-06-01

Smart wearables, known as emerging paradigms for vital big data capturing, have been attracting intensive attentions. However, one crucial problem is their power-hungriness, i.e., the continuous data streaming consumes energy dramatically and requires devices to be frequently charged. Targeting this obstacle, we propose to investigate the biodynamic patterns in the data and design a data-driven approach for intelligent data compression. We leverage Deep Learning (DL), more specifically, Convolutional Autoencoder (CAE), to learn a sparse representation of the vital big data. The minimized energy need, even taking into consideration the CAE-induced overhead, is tremendously lower than the original energy need. Further, compared with state-of-the-art wavelet compression-based method, our method can compress the data with a dramatically lower error for a similar energy budget. Our experiments and the validated approach are expected to boost the energy efficiency of wearables, and thus greatly advance ubiquitous big data applications in era of smart health.<br><div>In recent years, there has also been a growing interest in edge intelligence for emerging instantaneous big data inference. However, the inference algorithms, especially deep learning, usually require heavy computation requirements, thereby greatly limiting their deployment on the edge. We take special interest in the smart health wearable big data mining and inference. <br></div><div><br></div><div>Targeting the deep learning’s high computational complexity and large memory and energy requirements, new approaches are urged to make the deep learning algorithms ultra-efficient for wearable big data analysis. We propose to leverage knowledge distillation to achieve an ultra-efficient edge-deployable deep learning model. More specifically, through transferring the knowledge from a teacher model to the on-edge student model, the soft target distribution of the teacher model can be effectively learned by the student model. Besides, we propose to further introduce adversarial robustness to the student model, by stimulating the student model to correctly identify inputs that have adversarial perturbation. Experiments demonstrate that the knowledge distillation student model has comparable performance to the heavy teacher model but owns a substantially smaller model size. With adversarial learning, the student model has effectively preserved its robustness. In such a way, we have demonstrated the framework with knowledge distillation and adversarial learning can, not only advance ultra-efficient edge inference, but also preserve the robustness facing the perturbed input.</div>

健康的重要性，人人皆知，但是知易行難，現代的人忙碌，卻也創造出更多不運動的理由。近年來因為手機的普及，讓資通業者看到的是個人對身體自主健康的需求，隨著智慧手環的關鍵技術發展漸趨成熟，讓智慧穿戴裝置成為生活的必需品，許多人都相信智慧穿戴裝置將是下一個令人期待的高成長產業，而健康管理與穿戴裝置的結合又一直是此類產品發展的重點。 本論文研究的目的是希望經由分析智慧穿戴市場的基本應用與產業現況，來探討穿戴裝置業者的產品在健康管理相關領域的市場應用範圍。研究的架構是藉由分析幾個智慧穿戴業者現行的商業模式圖並輔以價值主張的設計分析探討，找出研究個案的關鍵成功因素，本研究也根據評量分析作出結論與建議，達成在這個互聯網時代，研究機構，廠家跟消費者能精準管理健康，產業成長獲利，全民健康運動的三贏目的。

No description available.