Attentional lapses while individuals are engaged in activities can have critical effects on their performance. EEG sensors offer the potential to monitor brain activity and detect such lapses in-situ. However, advances in automatic detection are limited, notably due to the scarcity of validated EEG data for training. In this work, we explore the design space of lapses-of-attention detectors using EEG signals, framing it as a binary classification problem. We introduce an EEG dataset with two validated attention levels acquired through a controlled experiment (N = 24) involving reading tasks with and without auditory distractions. We evaluated fifteen detectors using three different EEG feature extraction techniques, and five classifier models. Models using filterbank-CSP features yielded the highest median per-participant detection accuracy of 96%. Limited-resource analyses further indicate the Beta frequency band is the most informative for attention detection, and highlight detection can be achieved with only four EEG channels. Taken together, our findings inform on the feasibility and design of automatic attention detection for brain-computer interfaces utilizing simpler EEG devices.

Optical-see-through (OST) augmented reality headsets offer users the flexibility to access relevant data visualizations anytime and anywhere. However, the appearance of content displayed on OST displays varies in colour and transparency depending on the environment they are viewed in, potentially leading to interpretation challenges. We present the findings of a psychophysical study (N = 24), aimed at assessing the impact of two environmental factors – lighting intensity and background colour – on user performance and colour perception accuracy in a visualization and colour-matching task using an OST headset. Our results suggest the effect of background colour on visualization interpretation is notable only under bright lighting conditions. Interestingly, participants perceived low-colour-contrast scenarios as more challenging, although their performance did not decline. Additionally, visualization colours were perceptibly and distinctly mismatched, but did not blend with the background colours. Finally, we discuss visual comfort and colour coding in the context of designing ubiquitous visualizations on OST displays, highlighting open challenges.

Large curved displays are becoming increasingly popular due to their ability to provide users with a wider field of view and a more immersive experience compared to flat displays. Current interaction techniques for large curved displays often assume a user is positioned at the display's centre, crucially failing to accommodate general use conditions where the user may move during use. In this work, we investigated how user position impacts pointing interaction on large curved displays and evaluated cursor enhancement techniques to provide faster and more accurate performance across positions. To this effect, we conducted two user studies. First, we evaluated the effects of user position on pointing performance on a large semi-circular display (3m-tall, 3270R curvature) through a 2D Fitts' Law selection task. Our results indicate that as users move away from the display, their pointing speed significantly increases (at least by 9%), but accuracy decreases (by at least 6%). Additionally, we observed participants were slower when pointing from laterally offset positions. Secondly, we explored which pointing techniques providing motor- and visual-space enhancements best afford effective pointing performance across user positions. Across a total of six techniques tested, we found that a combination of acceleration and distance-based adjustments with cursor enlargement significantly improves target selection speed and accuracy across different user positions. Results further show techniques with visual-space enhancements (e.g., cursor enlargement) are significantly faster and more accurate than their non-visually-enhanced counterparts. Based on our results we provide design recommendations for implementing cursor enhancement techniques for large curved displays.

Conditionally automated vehicles provide unique opportunities for drivers to engage in non-driving-related tasks (NDRTs); however, drivers must remain prepared to respond to take-over requests. This paper explores design challenges and potential solutions for supporting reading as an NDRT in SAE Level 3 vehicles. Specifically, we assess two prominent text-presentation techniques: vertical scrolling text presentation (VSTP) and rapid serial visual presentation (RSVP), exploring both in conjunction with their integration with auditory speech displays (ASD). A driving simulation study involving N = 32 participants revealed that RSVP surpassed VSTP in regaining situational awareness, as indicated by lower average braking actuation, and was also preferred by participants. The integration of ASDs with both techniques reduced perceived cognitive workload and improved the user experience, albeit with compromised lateral control. Our findings can help advance the design of human-centered interfaces for reading in conditionally automated vehicles.

We combine ideas from shock graph theory with more recent appearance-based methods for medial axis extraction from complex natural scenes, improving upon the present best unsupervised method, in terms of efficiency and performance. We make the following specific contributions: i) we extend the shock graph representation to the domain of real images, by generalizing the shock type definitions using local, appearance-based criteria; ii) we then use the rules of a Shock Grammar to guide our search for medial points, drastically reducing run time when compared to other methods, which exhaustively consider all points in the input image; iii) we remove the need for typical post-processing steps including thinning, non-maximum suppression, and grouping, by adhering to the Shock Grammar rules while deriving the medial axis solution; iv) finally, we raise some fundamental concerns with the evaluation scheme used in previous work and propose a more appropriate alternative for assessing the performance of medial axis extraction from scenes. Our experiments on the BMAX500 and SK-LARGE datasets demonstrate the effectiveness of our approach. We outperform the present state-of-the-art, excelling particularly in the high-precision regime, while running an order of magnitude faster and requiring no post-processing.

We present a study of recent advances in computer vision (CV) research for the Global South to identify the main uses of modern CV and its most significant ethical risks in the region. We review 55 research papers and analyze them along three principal dimensions: where the technology was designed, the needs addressed by the technology, and the potential ethical risks arising following deployment. Results suggest: 1) CV is most used in policy planning and surveillance applications, 2) privacy violations is the most likely and most severe risk to arise from modern CV systems designed for the Global South, and 3) researchers from the Global North differ from researchers from the Global South in their uses of CV to solve problems in the Global South. Results of our risk analysis also differ from previous work on CV risk perception in the West, suggesting locality to be a critical component of each risk's importance.

Spike detection is a signal processing technique that can enable significant data rate reduction and resource savings in wireless brain monitoring. In these systems, energy-efficient spike detection algorithms are sought for enabling realtime signal processing while consuming low-power. As several spike detectors are based on ASIC, FPGA or low-power microcontroller unit (MCU), such algorithms must add little overhead to the entire system, while ensuring low error rate. In this paper, we present a comparative study of three different spike detection algorithms targeted toward implementation into low-power resource-constrained electronic systems. As practical validation, all candidate algorithms have been implemented on a popular low-power MCU and were fully characterized experimentally using previously recorded neural signals with different signal-to-noise ratios. A cost function based on detection rates, execution times, power consumption and resource utilization have been created and employed for comparing the detectors. The performances of all candidates are reported, and the best detector is identified. All candidate detectors present detection rate above 95% at high SNR, and above 78% for low SNR and can reduce the power consumption by up to 22.7%. This paper is the first to demonstrate the performances and hardware limitations of spike detectors on a low-power MCU system.

In this paper, we present a light-weight, wireless optogenetic headstage which provides optical neural stimulation and electrophysiological recording alongside on-the-fly neural signal processing. The proposed headstage is suitable to conduct long terms in-vivo experiments with small freely moving transgenic rodents, and features two implantable LED-coupled optical fibers and two electrophysiological recording channels while being powered by a small Lithium-ion battery. The headstage can transmit the raw neuronal signals or only spike waveforms after applying on-the-fly spike detection, which reduces power consumption by up to 14.5%. The headstage is entirely built using commercial off-the-shelf components, and the miniature design, using rigid-flex PCBs, results into a lightweight (7.4g) and compact device (25×20×15 mm). Low-power consumption is achieved by using on-the-fly spike detection alongside a real-time operating system which brings the headstage autonomy to 3h25 in full operation, including high-output power optical stimulation, micro-volts neuronal signal amplification and wireless transmission of the acquired waveforms.