This study investigates response time as a behavioral indicator related to listening effort (LE) for evaluating speech enhancement (SE) systems. English and Norwegian intelligibility matrix tests were conducted within a single-task paradigm that incorporated click-time recording (logging the precise time of all participant clicks), enabling simultaneous estimation of speech intelligibility and LE-related temporal behavior. Three temporal proxy measures for LE were examined—time per stimulus, reaction time, and word click time—across a broad range of input signal-to-noise ratios (SNRs) and for both discriminative and generative enhancement approaches. Time per stimulus showed an inverted-U pattern across SNRs, whereas reaction time and word click time exhibited monotonic behavior, providing more directly interpretable metrics for comparative evaluation. Analyses of pairwise SNR comparisons revealed that increases in our LE-related temporal measures at higher SNRs precede measurable intelligibility declines, suggesting that these temporal metrics can be more sensitive than intelligibility in this regime. Overall, the proposed framework—where LE-related measurements remain unknown to participants—offers a comprehensive and nuanced behavioral tool for SE evaluation, complementing intelligibility particularly under realistic, moderate-to-high SNR conditions.
Wireless Capsule Endoscopy (WCE) has emerged as a non-invasive and patient-friendly imaging modality for comprehensive visualization of the Gastrointestinal (GI) tract. However, due to the miniaturized capsule design, limited onboard optics, sensor constraints, and wireless transmission bandwidth restrictions, WCE images are inherently acquired at low spatial resolution, often accompanied by noise, motion blur, and illumination degradation. These limitations significantly reduce the visibility of fine anatomical structures such as mucosal textures, vascular patterns, and lesion boundaries, thereby affecting diagnostic reliability. To address this challenge, this paper proposes a computationally efficient unsupervised Transformer-based super-resolution framework, termed as CEM-TUDASR, for enhancing WCE images without relying on paired Low-Resolution (LR) and High-Resolution (HR) training data. The proposed framework integrates a domain-adaptive degradation modeling network that learns to synthesize realistic WCE-like LR images from HR conventional endoscopy images, enabling effective unpaired training and reducing the domain discrepancy between conventional and capsule endoscopy data. Furthermore, a Transformer-based SR generator incorporating Deep Attention Blocks (DABs) and a Fusion Attention Block (FAB) is introduced to jointly capture long-range contextual dependencies and fine-grained local structural details. This architecture facilitates improved reconstruction of diagnostically relevant regions while preserving structural consistency and perceptual fidelity. The proposed model is trained on a newly curated WCE dataset derived from the Kvasir Capsule dataset and extensively evaluated on external benchmark datasets, including KID and GIANA, to validate its effectiveness and generalization capability. Quantitative evaluation using no-reference image quality assessment metrics, including BRISQUE, PIQE, NIQE, and the domain-specific EndoQM, demonstrates that the proposed method consistently outperforms existing unsupervised SR approaches in terms of perceptual quality, structural preservation, and domain-specific visual fidelity. Qualitative analysis further confirms superior restoration of subtle mucosal textures, vascular structures, and clinically significant anatomical details essential for accurate interpretation. In addition, cross-domain evaluation on retinal images demonstrates the robustness and adaptability of the proposed framework across diverse medical imaging modalities. Despite achieving high-quality reconstruction performance, the proposed architecture maintains computational efficiency with only 2.67 million parameters and 169.94 GFLOPs, making it highly suitable for deployment in real-time and resource-constrained clinical environments, including portable and embedded endoscopic systems. GitHub Link: https://github.com/Jay042003/CEM-TUDASR.git.
Optimized future floating wind turbines (FWTs) are expected to be both larger and relatively lighter than conventional offshore platforms, thus more flexible. The common practice of modeling the platform as a rigid body in coupled dynamic simulations of FWTs can then be questioned. Specifically, natural frequencies of the elastic modes of large flexible platforms can be close to the frequency range of excitation loads. Additionally, platform flexibility can have a significant effect on the natural modes involving significant tower deformation. Considering the platform’s flexibility in coupled simulations of large FWTs requires distributing the hydrodynamic and hydrostatic pressure loads on the flexible model of the platform, instead of the traditional approach of lumping the loads at a single point. This work presents a rational method to evaluate the first-order added mass, radiation damping, and excitation coefficients for a multi-body representation of the platform and develops an energy-conserving distributed formulation for the hydrostatic loads. Assuming small flexible deformations, a decoupled radiation damping matrix is used to model radiation loads for better computational efficiency, while a fully coupled infinite-frequency added mass matrix is used to ensure a stable model. The decoupled radiation coefficients can either be obtained from a single-body or a multi-body diffraction/radiation analysis. A case study of the INO OptiFLEX 22MW semisubmersible FWT is used to illustrate and verify the implementation of the proposed approach. Compared to a baseline model with rigid floater, the results show that introducing platform flexibility significantly affects the high-frequency dynamics of the tower and can also potentially affect mooring line tensions. Moreover, platform flexibility was shown to influence roll and pitch dynamics. These findings highlight the need to model platform flexibility in coupled simulations when analyzing and designing future large FWTs, which can be achieved through the proposed methodology.
Software beamforming has enabled the introduction of a myriad adaptive beamformers. To reproduce and compare those techniques, a unified beamforming framework is imperative. Here, we propose such a framework, called the Generalized Beamformer (GB), which provides a unified approach to most state-of-the-art beamforming techniques through a single core expression. This generalization is achieved through a range of delay and apodization models that enable translation across various transmit sequences. Exploiting the GB we also demonstrate a novel synthetic transmit focusing strategy for adaptive beamforming, where signals are coherently combined across transmit events prior to e.g. Coherence Factor or Minimum Variance processing. For the investigated CPWC case, this substantially reduces computational complexity while improving resolution and maintaining contrast. The GB is implemented within the open-source UltraSound ToolBox (USTB) for MATLAB, and its pixel-based approach provides flexibility in scan grid definition. The framework is shown to successfully beamform conventional transmit sequences, including focused, diverging, plane, and single-element transmissions. Its versatility is showcased through diverse applications, including in-vivo cardiac and fetal imaging, as well as subsea sonar.
Conventional all-air cooling systems struggle with high latent loads and increased energy demand in tropical buildings, where humidity and solar radiation intensify the design challenge. This study systematically reviews radiant cooling system (RCS) research in tropical and hot, humid climates, synthesizing evidence from 152 peerreviewed studies published between 2013 and 2025. The review is structured around four dimensions: operating principles and system types, energy efficiency, thermal comfort, and advanced control strategies. Key quantitative findings include: properly designed RCSs reduce energy consumption by 4-43% relative to conventional air systems, with peak savings of up to 56% when integrated with phase change materials or geothermal heat pumps; RCSs maintain comfort at air temperatures 1-2 K higher than traditional systems, with a predicted mean vote within the acceptable range (-0.5 to +0.5) for over 81% of occupied hours; model predictive control maintains EN 15251 Category II standards for more than 95% of occupied hours while reducing cooling-tower energy use by up to 55%, with membrane-assisted sub-dew-point panels reducing annual discomfort hours by 3-6% versus conventional RCS. The primary contribution is a structured, evidence-based framework that: (1) taxonomizes hybrid RCS configurations by tropical microclimate severity; (2) synthesizes condensation-risk thresholds and control design criteria; and (3) delineates a technology-readiness roadmap for net-zero buildings in humid environments. Research gaps include the absence of in situ validation of membrane-assisted RCS in operational tropical buildings, the lack of standardized condensation-risk thresholds, and the need for machine learning controllers validated across diverse tropical microclimates.
