264 papers
This paper introduces Deep Unfolded Full Waveform Inversion (DUFWI), a physics-informed deep learning method that achieves real-time, high-quality speed-of-sound reconstruction for edema detection by overcoming the computational intensity and local minima issues of traditional inversion techniques.
This paper characterizes the frequency response of windowed DFT-based phasor estimators used in PMUs, deriving their complex gain to explain how estimation windows distort oscillation measurements and proposing a method to recover true oscillation parameters for improved observability.
This paper proposes a balanced 6G waveform strategy that integrates orthogonality and non-orthogonality, identifying SC-NOFS(2D) as a superior candidate over SC-OFDM for achieving high data rates, low latency, and robust mobility while ensuring backward compatibility.
This paper proposes a parameter-free activation steering method that identifies accent information within a specific band of middle encoder layers in speech foundation models and corrects accent-induced representation shifts during inference, thereby significantly reducing word error rates across diverse accents without requiring model fine-tuning.
ImKWS is a novel test-time adaptation method for keyword spotting that addresses severe class imbalance between rare keywords and background noise by employing a dual-branch entropy minimization strategy with separate update strengths and multi-transformation consistency, thereby preventing model overconfidence and bias without requiring labeled data.
This article reviews the state-of-the-art of stacked intelligent metasurfaces (SIMs) for in-wave computation in next-generation networks, highlighting their potential to jointly optimize communication, sensing, and processing while addressing current challenges in scalability, controllability, and robustness.
This paper analyzes the ill-posedness of CSI-based electromagnetic inverse scattering in ISAC systems by characterizing the coherence of scattering operators, proving that restricting the reconstruction to a Region of Interest (ROI) significantly improves conditioning and estimation bounds, and validating a corresponding ROI-constrained optimization framework through simulations.
This paper introduces a training-free, post-hoc method called activation steering that neutralizes accents in zero-shot Text-to-Speech while preserving speaker timbre by applying offline-extracted steering vectors during inference.
This paper proposes a conditioning-aware channel estimation framework for RIS-assisted upper mid-band MIMO systems that overcomes severe spatial correlation and ill-conditioning in near-field propagation by transforming the high-dimensional problem into well-conditioned subproblems through greedy column grouping and piecewise phase design, thereby achieving robust performance without relying on sparsity assumptions.
This paper proposes a model-based reinforcement learning strategy to optimize relay power allocation in an underwater hybrid optical-acoustic energy harvesting network, effectively maximizing long-term secrecy performance against acoustic eavesdropping while adapting to dynamic channel conditions and battery constraints.
This paper proposes a unified retrieval-assisted framework for wireless localization that combines channel charting for efficient low-dimensional retrieval of reference points with a graph attention network to model inter-sample correlations, thereby achieving superior accuracy and scalability compared to existing similarity-based and learning-based methods.
The paper introduces MAD, a novel multimodal affective dataset featuring synchronized physiological signals (EEG, ECG, EOG, EMG, PPG, BCG) and tri-view RGB-D facial videos from 18 participants, accompanied by a hierarchical three-level annotation framework to facilitate comprehensive research in emotion recognition and cross-modal affective analysis.
This paper introduces a novel set-prediction-based framework for J-peak detection in pillow-based ballistocardiography, supported by a newly released multi-subject dataset, which outperforms traditional segmentation approaches while significantly reducing model complexity and eliminating the need for heuristic post-processing.
This cross-linguistic study demonstrates that while certain acoustic-prosodic markers of autism in children's speech generalize across Finnish, French, and Slovak, robust classification performance requires language-specific modeling due to significant variations in feature importance and transferability across typologically distinct languages.
This study analyzes a multilingual corpus of Finnish, French, and Slovak child speech to demonstrate that autistic speakers exhibit a distinct, cross-linguistic prosodic profile characterized by increased intensity variability, clearer voice quality, and reduced temporal dynamics, thereby challenging deficiency-based models in favor of a complex, language-independent acoustic signature.
This paper presents a competitive open-source cascaded system that combines EEND-VC speaker diarization and fine-tuned Qwen3 ASR to achieve first place in the DISPLACE-M challenge by effectively extracting medical conditions from overlapping, code-switched Hinglish clinical dialogues.
This paper proposes a unified Lagrangian regression framework that naturally incorporates the Discrete Cosine Transform (DCT) as a novel, computationally efficient, and convergent alternative to traditional polynomial and logistic regression methods.
This paper proposes a unified multicarrier waveform framework for 6G networks by systematically characterizing and comparing state-of-the-art one-dimensional and two-dimensional modulation schemes to guide their selection based on channel conditions and performance requirements.
This paper investigates the impact of parameter selection on the convergence of a modified ant colony algorithm to optimize the hierarchical structure of an industrial distributed control system, offering tuning recommendations applicable to broader combinatorial optimization problems.
This paper investigates the use of a text-trained large language model adapted for speech to localize fake words in partially edited audio, revealing that while the model effectively identifies edits by leveraging specific training patterns like word-level polarity substitutions, it struggles to generalize to unseen editing styles.