Optimal input settings vary across users due to differences in motor abilities and personal preferences, which are typically addressed by manual tuning or calibration. Although human-in-the-loop optimization has the potential to identify optimal settings during use, it is rarely applied due to its long optimization process. A more efficient approach would continually leverage data from previous users to accelerate optimization, exploiting shared traits while adapting to individual characteristics. We introduce the concept of Continual Human-in-the-Loop Optimization and a Bayesian optimization-based method that leverages a Bayesian-neural-network surrogate model to capture population-level characteristics while adapting to new users. We propose a generative replay strategy to mitigate catastrophic forgetting. We demonstrate our method by optimizing virtual reality keyboard parameters for text entry using direct touch, showing reduced adaptation times with a growing user base. Our method opens the door for next-generation personalized input systems that improve with accumulated experience.

Touch contact and pressure are essential for understanding how humans interact with objects and offer insights that benefit applications in mixed reality and robotics. Estimating these interactions from an egocentric camera perspective is challenging, largely due to the lack of comprehensive datasets that provide both hand poses and pressure annotations. In this paper, we present EgoPressure, an egocentric dataset that is annotated with high-resolution pressure intensities at contact points and precise hand pose meshes, obtained via our multi-view, sequence-based optimization method. We introduce baseline models for estimating applied pressure on external surfaces from RGB images, both with and without hand pose information, as well as a joint model for predicting hand pose and the pressure distribution across the hand mesh. Our experiments show that pressure and hand pose complement each other in understanding hand-object interactions.

Motion capture using sparse inertial sensors has shown great promise due to its portability and lack of occlusion issues compared to camera-based tracking. Existing approaches typically assume that IMU sensors are tightly attached to the human body. However, this assumption often does not hold in real-world scenarios. In this paper, we present a new task of full-body human pose estimation using sparse, loosely attached IMU sensors. To solve this task, we simulate IMU recordings from an existing garment-aware human motion dataset. We developed transformer-based diffusion models to synthesize loose IMU data and estimate human poses based on this challenging loose IMU data. In addition, we show that incorporating garment-related parameters while training the model on simulated loose data effectively maintains expressiveness and enhances the ability to capture variations introduced by looser or tighter garments. Experiments show that our proposed diffusion methods trained on simulated and synthetic data outperformed the state-of-the-art methods quantitatively and qualitatively, opening up a promising direction for future research.

Understanding affect is central to anticipating human behavior, yet current egocentric vision benchmarks largely ignore the person’s emotional states that shape their decisions and actions. Existing tasks in egocentric perception focus on physical activities, hand-object interactions, and attention modeling—assuming neutral affect and uniform personality. This limits the ability of vision systems to capture key internal drivers of behavior. In this paper, we present egoEMOTION, the first dataset that couples egocentric visual and physiological signals with dense self-reports of emotion and personality across controlled and real-world scenarios. Our dataset includes over 50 hours of recordings from 43 participants, captured using Meta’s Project Aria glasses. Each session provides synchronized eye-tracking video, head-mounted photoplethysmography, inertial motion data, and physiological baselines for reference. Participants completed emotion-elicitation tasks and naturalistic activities while self-reporting their affective state using the Circumplex Model and Mikels’ Wheel as well as their personality via the Big Five model. We define three benchmark tasks: (1) continuous affect classification (valence, arousal, dominance); (2) discrete emotion classification; and (3) trait-level personality inference. We show that a classical learning-based method, as a simple baseline in real-world affect prediction, produces better estimates from signals captured on egocentric vision systems than processing physiological signals. Our dataset establishes emotion and personality as core dimensions in egocentric perception and opens new directions in affect-driven modeling of behavior, intent, and interaction.

Research on egocentric tasks in computer vision has mostly focused on head-mounted cameras, such as fisheye cameras or embedded cameras inside immersive headsets.We argue that the increasing miniaturization of optical sensors will lead to the prolific integration of cameras into many more body-worn devices at various locations.This will bring fresh perspectives to established tasks in computer vision and benefit key areas such as human motion tracking, body pose estimation, or action recognition---particularly for the lower body, which is typically occluded.In this paper, we introduce EgoSim, a novel simulator of body-worn cameras that generates realistic egocentric renderings from multiple perspectives across a wearer's body.A key feature of EgoSim is its use of real motion capture data to render motion artifacts, which are especially noticeable with arm- or leg-worn cameras.In addition, we introduce MultiEgoView, a dataset of egocentric footage from six body-worn cameras and ground-truth full-body 3D poses during several activities:119 hours of data are derived from AMASS motion sequences in four high-fidelity virtual environments, which we augment with 5 hours of real-world motion data from 13 participants using six GoPro cameras and 3D body pose references from an Xsens motion capture suit.We demonstrate EgoSim's effectiveness by training an end-to-end video-only 3D pose estimation network.Analyzing its domain gap, we show that our dataset and simulator substantially aid training for inference on real-world data.EgoSim code & MultiEgoView dataset: https://siplab.org/projects/EgoSim

Mixed Reality is increasingly used in mobile settings beyond controlled home and office spaces. This mobility introduces the need for user interface layouts that adapt to varying contexts. However, existing adaptive systems are designed only for static environments. In this paper, we introduce SituationAdapt, a system that adjusts Mixed Reality UIs to real-world surroundings by considering environmental and social cues in shared settings. Our system consists of perception, reasoning, and optimization modules for UI adaptation. Our perception module identifies objects and individuals around the user, while our reasoning module leverages a Vision-and-Language Model to assess the placement of interactive UI elements. This ensures that adapted layouts do not obstruct relevant environmental cues or interfere with social norms. Our optimization module then generates Mixed Reality interfaces that account for these considerations as well as temporal constraints. For evaluation, we first validate our reasoning module’s capability of assessing UI contexts in comparison to human expert users. In an online user study, we then establish SituationAdapt’s capability of producing context-aware layouts for Mixed Reality, where it outperformed previous adaptive layout methods. We conclude with a series of applications and scenarios to demonstrate SituationAdapt’s versatility.

While passive surfaces offer numerous benefits for interaction in mixed reality, reliably detecting touch input solely from head-mounted cameras has been a long-standing challenge. Camera specifics, hand self-occlusion, and rapid movements of both head and fingers introduce considerable uncertainty about the exact location of touch events. Existing methods have thus not been capable of achieving the performance needed for robust interaction.

Current Virtual Reality systems are designed for interaction under visual control. Using built-in cameras, headsets track the user’s hands or hand-held controllers while they are inside the field of view. Current systems thus ignore the user’s interaction with off-screen content—virtual objects that the user could quickly access through proprioception without requiring laborious head motions to bring them into focus. In this paper, we present HOOV, a wrist-worn sensing method that allows VR users to interact with objects outside their field of view. Based on the signals of a single wrist-worn inertial sensor, HOOV continuously estimates the user’s hand position in 3-space to complement the headset’s tracking as the hands leave the tracking range. Our novel data-driven method predicts hand positions and trajectories from just the continuous estimation of hand orientation, which by itself is stable based solely on inertial observations. Our inertial sensing simultaneously detects finger pinching to register off-screen selection events, confirms them using a haptic actuator inside our wrist device, and thus allows users to select, grab, and drop virtual content. We compared HOOV’s performance with a camera-based optical motion capture system in two folds. In the first evaluation, participants interacted based on tracking information from the motion capture system to assess the accuracy of their proprioceptive input, whereas in the second, they interacted based on HOOV’s real-time estimations. We found that HOOV’s target-agnostic estimations had a mean tracking error of 7.7 cm, which allowed participants to reliably access virtual objects around their body without first bringing them into focus. We demonstrate several applications that leverage the larger input space HOOV opens up for quick proprioceptive interaction, and conclude by discussing the potential of our technique.

We present a novel learning-based method that achieves state-of-the-art performance on several heart rate estimation benchmarks extracted from photoplethysmography signals (PPG). We consider the evolution of the heart rate in the context of a discrete-time stochastic process that we represent as a hidden Markov model. We derive a distribution over possible heart rate values for a given PPG signal window through a trained neural network. Using belief propagation, we incorporate the statistical distribution of heart rate changes to refine these estimates in a temporal context. From this, we obtain a quantized probability distribution over the range of possible heart rate values that captures a meaningful and well-calibrated estimate of the inherent predictive uncertainty. We show the robustness of our method on eight public datasets with three different cross-validation experiments.

Despite advancements in egocentric hand tracking using head-mounted cameras, contact detection with real-world objects remains challenging, particularly for the quick motions often performed during interaction in Mixed Reality. In this paper, we introduce a novel method for detecting touch on discovered physical surfaces purely from an egocentric perspective using optical sensing. We leverage structured laser light to detect real-world surfaces from the disparity of reflections in real-time and, at the same time, extract a time series of remote vibrometry sensations from laser speckle motions. The pattern caused by structured laser light reflections enables us to simultaneously sample the mechanical vibrations that propagate through the user’s hand and the surface upon touch.

Teleconferencing is poised to become one of the most frequent use cases of immersive platforms, since it supports high levels of presence and embodiment in collaborative settings. On desktop and mobile platforms, teleconferencing solutions are already among the most popular apps and accumulate significant usage time---not least due to the pandemic or as a desirable substitute for air travel or commuting. In this paper, we present ViGather, an immersive teleconferencing system that integrates users of all platform types into a joint experience via equal representation and a first-person experience. ViGather renders all participants as embodied avatars in one shared scene to establish co-presence and elicit natural behavior during collocated conversations, including nonverbal communication cues such as eye contact between participants as well as body language such as turning one's body to another person or using hand gestures to emphasize parts of a conversation during the virtual hangout. Since each user embodies an avatar and experiences situated meetings from an egocentric perspective no matter the device they join from, ViGather alleviates potential concerns about self-perception and appearance while mitigating potential 'Zoom fatigue', as users' self-views are not shown. For participants in Mixed Reality, our system leverages the rich sensing and reconstruction capabilities of today's headsets. For users of tablets, laptops, or PCs, ViGather reconstructs the user's pose from the device's front-facing camera, estimates eye contact with other participants, and relates these non-verbal cues to immediate avatar animations in the shared scene. Our evaluation compared participants' behavior and impressions while videoconferencing in groups of four inside ViGather with those in Meta Horizon as a baseline for a social VR setting. Participants who participated on traditional screen devices (e.g., laptops and desktops) using ViGather reported a significantly higher sense of physical, spatial, and self-presence than when using Horizon, while all perceived similar levels of active social presence when using Virtual Reality headsets. Our follow-up study confirmed the importance of representing users on traditional screen devices as reconstructed avatars for perceiving self-presence.

Despite the advent of touchscreens, typing on physical keyboards remains most efficient for entering text, because users can leverage all fingers across a full-size keyboard for convenient typing. As users increasingly type on the go, text input on mobile and wearable devices has had to compromise on full-size typing. In this paper, we present TapType, a mobile text entry system for full-size typing on passive surfaces—without an actual keyboard. From the inertial sensors inside a band on either wrist, TapType decodes and relates surface taps to a traditional QWERTY keyboard layout. The key novelty of our method is to predict the most likely character sequences by fusing the finger probabilities from our Bayesian neural network classifier with the characters’ prior probabilities from an n-gram language model. In our online evaluation, participants on average typed 19 words per minute with a character error rate of 0.6% after 30 minutes of training. Expert typists thereby consistently achieved more than 25 WPM at a similar error rate. We demonstrate applications of TapType in mobile use around smartphones and tablets, as a complement to interaction in situated Mixed Reality outside visual control, and as an eyes-free mobile text input method using an audio feedback-only interface.

Real-world work-spaces typically revolve around tables, which enable knowledge workers to comfortably perform tasks over an extended period of time during productivity tasks. Tables afford more ergonomic postures and provide opportunities for rest, which raises the question of whether they may also benefit prolonged interaction in Virtual Reality (VR). In this paper, we investigate the effects of tabletop surface presence in situated VR settings on task performance, behavior, and subjective experience. In an empirical study, 24 participants performed two tasks (selection, docking) on virtual interfaces placed at two distances and two orientations. Our results show that a physical tabletop inside VR improves comfort, agency, and task performance while decreasing physical exertion and strain of the neck, shoulder, elbow, and wrist, assessed through objective metrics and subjective reporting. Notably, we found that these benefits apply when the UI is placed on and aligned with the table itself as well as when it is positioned vertically in mid-air above it. Our experiment therefore provides empirical evidence for integrating physical table surfaces into VR scenarios to enable and support prolonged interaction. We conclude by discussing the effective usage of surfaces in situated VR experiences and provide initial guidelines.

Touch input is dominantly detected using mutual-capacitance sensing, which measures the proximity of close-by objects that change the electric field between the sensor lines. The exponential drop-off in intensities with growing distance enables software to detect touch events, but does not reveal true contact areas. In this paper, we introduce CapContact, a novel method to precisely infer the contact area between the user’s finger and the surface from a single capacitive image. At 8 × super-resolution, our convolutional neural network generates refined touch masks from 16-bit capacitive images as input, which can even discriminate adjacent touches that are not distinguishable with existing methods. We trained and evaluated our method using supervised learning on data from 10 participants who performed touch gestures. Our capture apparatus integrates optical touch sensing to obtain ground-truth contact through high-resolution frustrated total internal reflection. We compare our method with a baseline using bicubic upsampling as well as the ground truth from FTIR images. We separately evaluate our method’s performance in discriminating adjacent touches. CapContact successfully separated closely adjacent touch contacts in 494 of 570 cases (87%) compared to the baseline’s 43 of 570 cases (8%). Importantly, we demonstrate that our method accurately performs even at half of the sensing resolution at twice the grid-line pitch across the same surface area, challenging the current industry-wide standard of a ∼ 4 mm sensing pitch. We conclude this paper with implications for capacitive touch sensing in general and for touch-input accuracy in particular.

Today’s touchscreen devices commonly detect the coordinates of user input through capacitive sensing. Yet, these coordinates are the mere 2D manifestations of the more complex 3D configuration of the whole hand—a sensation that touchscreen devices so far remain oblivious to. In this work, we introduce the problem of reconstructing a 3D hand skeleton from capacitive images, which encode the sparse observations captured by touch sensors. These low-resolution images represent intensity mappings that are proportional to the distance to the user’s fingers and hands.

Current Virtual Reality systems typically use cameras to capture user input from controllers or free-hand mid-air interaction. In this paper, we argue that this is a key impediment to productivity scenarios in VR, which require continued interaction over prolonged periods of time-a requirement that controller or free-hand input in mid-air does not satisfy. To address this challenge, we bring rapid touch interaction on surfaces to Virtual Reality-the input modality that users have grown used to on phones and tablets for continued use. We present TapID, a wrist-based inertial sensing system that complements headset-tracked hand poses to trigger input in VR. TapID embeds a pair of inertial sensors in a flexible strap, one at either side of the wrist; from the combination of registered signals, TapID reliably detects surface touch events and, more importantly, identifies the finger used for touch. We evaluated TapID in a series of user studies on event-detection accuracy (F1 = 0.997) and hand-agnostic finger-identification accuracy (within-user: F1 = 0.93; across users: F1 = 0.91 after 10 refinement taps and F1 = 0.87 without refinement) in a seated table scenario. We conclude with a series of applications that complement hand tracking with touch input and that are uniquely enabled by TapID, including UI control, rapid keyboard typing and piano playing, as well as surface gestures.