Papers

In this paper, we investigate the uplink (UL) radio resource management for 5G aerial corridors with an open-radio access network (O-RAN)-enabled cell-free (CF) massive multiple-input multiple-output (mMIMO) system. Our objective is to maximize the minimum spectral efficiency (SE) by jointly optimizing unmanned aerial vehicle (UAV)-open radio unit (O-RU) association and UL transmit power under quality-of-service (QoS) constraints. Owing to its NP-hard nature, the formulated problem is decomposed into two tractable sub-problems solved via alternating optimization (AO) using two computationally efficient algorithms. We then propose (i) a QoS-driven and multi-connectivity-enabled association algorithm incorporating UAV-centric and O-RU-centric criteria with targeted refinement for weak UAVs, and (ii) a bisection-guided fixed-point power control algorithm achieving global optimality with significantly reduced complexity, hosted as xApp at the near-real-time (near-RT) RAN intelligent controller (RIC) of O-RAN. Solving the resource-allocation problem requires global channel state information (CSI), which incurs substantial measurement and signaling overhead. To mitigate this, we leverage a channel knowledge map (CKM) within the O-RAN non-RT RIC to enable efficient environment-aware CSI inference. Simulation results show that the proposed framework achieves up to 440% improvement in minimum SE, 100% QoS satisfaction and fairness, while reducing runtime by up to 99.7% compared to an interior point solver-based power allocation solution, thereby enabling O-RAN compliant real-time deployment.

This paper addresses joint admission control and per-user equipment (UE) bandwidth allocation to maximize weighted sum-rate in network slicing-enabled user-centric cell-free (CF) massive multiple-input multiple-output (mMIMO) systems when aggregate quality-of-service (QoS) demand may exceed available bandwidth. Specifically, we optimize bandwidth allocation while satisfying heterogeneous QoS requirements across enhanced mobile broadband (eMBB) and ultra-reliable low-latency communication (URLLC) slices in the uplink. The formulated problem is NP-hard, rendering global optimality computationally intractable. We decompose it into two sub-problems and solve them via computationally efficient heuristics within a sequential framework. We propose (i) a hierarchical admission control scheme that selectively admits UEs under bandwidth scarcity, prioritizing URLLC to ensure latency-sensitive QoS compliance, and (ii) an iterative gradient-based bandwidth allocation scheme that transfers bandwidth across slices guided by marginal utility and reallocates resources within slices. Simulation results demonstrate that the proposed scheme achieves near-optimal performance, deviating from an interior point solver-based benchmark by at most 2.2% in weighted sum-rate while reducing runtime by 99.7%, thereby enabling practical real-time deployment. Compared to a baseline round-robin scheme without admission control, the proposed approach achieves up to 1085% and 7% higher success rates for eMBB and URLLC slices, respectively, by intentionally sacrificing sum-rate to guarantee QoS. Sensitivity analysis further reveals that the proposed solution adapts effectively to diverse eMBB/URLLC traffic compositions, maintaining 47-51% eMBB and 93-94% URLLC success rates across varying load scenarios, confirming its robustness for resource-constrained large-scale deployments.

This paper addresses the joint optimization of per-user equipment (UE) bandwidth allocation and UE-access point (AP) association to maximize weighted sum-rate while satisfying heterogeneous quality-of-service (QoS) requirements across enhanced mobile broadband (eMBB) and ultra-reliable low-latency communication (URLLC) slices in the uplink of a network slicing-enabled user-centric cell-free (CF) massive multiple-input multiple-output (mMIMO) system. The formulated problem is NP-hard, rendering global optimality computationally intractable. To address this challenge, it is decomposed into two sub-problems, each solved by a computationally efficient heuristic scheme, and jointly optimized through an alternating optimization framework. We then propose (i) a bandwidth allocation scheme that balances UE priority, spectral efficiency, and minimum bandwidth demand under limited resources to ensure fair QoS distribution, and (ii) a priority-based UE-AP association assignment approach that balances UE service quality with system capacity constraints. Together, these approaches provide a practical and computationally efficient solution for resource-constrained network slicing scenarios, where QoS feasibility is often violated under dense deployments and limited bandwidth, necessitating graceful degradation and fair QoS preservation rather than solely maximizing the aggregate sum-rate. Simulation results demonstrate that the proposed scheme achieves up to 52% higher weighted sum-rate, 140% and 58% higher QoS success rates for eMBB and URLLC slices, respectively, while reducing runtime by up to 97% compared to the considered benchmarks.

With the widespread application of 3D Gaussians in 3D scene representation, 3D scene segmentation methods based on 3D Gaussians have also gradually emerged. However, existing 3D Gaussian segmentation methods basically segment on the basis of Gaussian primitives. Due to the large variation range of the scale of 3D Gaussians, large-sized Gaussians that often span the foreground and background lead to jagged boundaries of segmented objects. To this end, we propose an online boundary trimming method, GaussianTrimmer, which is an efficient and plug-and-play post-processing method capable of trimming coarse boundaries for existing 3D Gaussian segmentation methods. Our method consists of two core steps: 1. Generating uniformly and well-covered virtual cameras; 2. Trimming Gaussian at the primitive level based on 2D segmentation results on virtual cameras. Extensive quantitative and qualitative experiments demonstrate that our method can improve the segmentation quality of existing 3D Gaussian segmentation methods as a plug-and-play method.

Traditional textile factories consume substantial energy, making energy-efficient production optimization crucial for sustainability and cost reduction. Meanwhile, deep neural networks (DNNs), which are effective for factory output prediction and operational optimization, require extensive historical data, posing challenges due to high sensor deployment and data collection costs. To address this, we propose Ensemble Deep Transfer Learning (EDTL), a novel framework that enhances prediction accuracy and data efficiency by integrating transfer learning with an ensemble strategy and a feature alignment layer. EDTL pretrains DNN models on data-rich production lines (source domain) and adapts them to data-limited lines (target domain), reducing dependency on large datasets. Experiments on real-world textile factory datasets show that EDTL improves prediction accuracy by 5.66% and enhances model robustness by 3.96% compared to conventional DNNs, particularly in data-limited scenarios (20%-40% data availability). This research contributes to energy-efficient textile manufacturing by enabling accurate predictions with fewer data requirements, providing a scalable and cost-effective solution for smart production systems.

We address real-time sampling and estimation of autoregressive Markovian sources in dynamic yet structurally similar multi-hop wireless networks. Each node caches samples from others and communicates over wireless collision channels, aiming to minimize time-average estimation error via decentralized policies. Due to the high dimensionality of action spaces and complexity of network topologies, deriving optimal policies analytically is intractable. To address this, we propose a graphical multi-agent reinforcement learning framework for policy optimization. Theoretically, we demonstrate that our proposed policies are transferable, allowing a policy trained on one graph to be effectively applied to structurally similar graphs. Numerical experiments demonstrate that (i) our proposed policy outperforms state-of-the-art baselines; (ii) the trained policies are transferable to larger networks, with performance gains increasing with the number of agents; (iii) the graphical training procedure withstands non-stationarity, even when using independent learning techniques; and (iv) recurrence is pivotal in both independent learning and centralized training and decentralized execution, and improves the resilience to non-stationarity.

Interpretability is essential for user trust in real-world anomaly detection applications. However, deep learning models, despite their strong performance, often lack transparency. In this work, we study the interpretability of autoencoder-based models for audio anomaly detection, by comparing a standard autoencoder (AE) with a mask autoencoder (MAE) in terms of detection performance and interpretability. We applied several attribution methods, including error maps, saliency maps, SmoothGrad, Integrated Gradients, GradSHAP, and Grad-CAM. Although MAE shows a slightly lower detection, it consistently provides more faithful and temporally precise explanations, suggesting a better alignment with true anomalies. To assess the relevance of the regions highlighted by the explanation method, we propose a perturbation-based faithfulness metric that replaces them with their reconstructions to simulate normal input. Our findings, based on experiments in a real industrial scenario, highlight the importance of incorporating interpretability into anomaly detection pipelines and show that masked training improves explanation quality without compromising performance.

Unmanned aerial vehicles (UAVs) are pivotal for future 6G non-terrestrial networks, yet their high mobility creates a complex coupled optimization problem for beamforming and trajectory design. Existing numerical methods suffer from prohibitive latency, while standard deep learning often ignores dynamic interference topology, limiting scalability. To address these issues, this paper proposes a hierarchically decoupled framework synergizing graph neural networks (GNNs) with multi-agent reinforcement learning. Specifically, on the short timescale, we develop a topology-aware GNN beamformer by incorporating GraphNorm. By modeling the dynamic UAV-user association as a time-varying heterogeneous graph, this method explicitly extracts interference patterns to achieve sub-millisecond inference. On the long timescale, trajectory planning is modeled as a decentralized partially observable Markov decision process and solved via the multi-agent proximal policy optimization algorithm under the centralized training with decentralized execution paradigm, facilitating cooperative behaviors. Extensive simulation results demonstrate that the proposed framework significantly outperforms state-of-the-art optimization heuristics and deep learning baselines in terms of system sum rate, convergence speed, and generalization capability.

Grid resilience is crucial in light of power interruptions caused by increasingly frequent extreme weather events. Well-designed energy management systems (EMS) have made progress in improving microgrid resilience through the coordination of distributed energy resources (DERs), but still face significant challenges in addressing the uncertainty of load demand caused by extreme weather. The integration of deep reinforcement learning (DRL) into EMS design enables optimized microgrid control strategies for coordinating DERs. Building on this, we proposed a cooperative multi-agent deep reinforcement learning (MADRL)-based EMS framework to provide flexible scalability for microgrids, enhance resilience and reduce operational costs during power outages. Specifically, the gated recurrent unit with a gating mechanism was introduced to extract features from temporal data, which enables the EMS to coordinate DERs more efficiently. Next, the proposed MADRL method incorporating action masking techniques was evaluated in the IEEE 33-Bus system using real-world data on renewable generation and power load. Finally, the numerical results demonstrated the superiority of the proposed method in reducing operating costs as well as the effectiveness in enhancing microgrid resilience during power interruptions.

Multi-beam radar sensing systems are emerging as powerful tools for non-contact motion tracking and vital-sign monitoring in healthcare environments. This paper presents the design and experimental validation of a synchronized millimeter-wave multi-radar tracking system enhanced by a modified spherical gradient-index (GRIN) Luneburg lens. Five commercial FMCW radar modules operating in the 58–63 GHz band are arranged in a semi-circular configuration around the lens, whose tailored refractive-index profile accommodates bistatic radar modules with co-located transmit (TX) and receive (RX) antennas. The resulting architecture generates multiple fixed high-gain beams with improved angular resolution and minimal mutual interference. Each radar operates independently but is temporally synchronized through a centralized Python-based acquisition framework to enable parallel data collection and low-latency motion tracking. A 10-cm-diameter 3D-printed prototype demonstrates a measured gain enhancement of approximately 12 dB for each module, corresponding to a substantial improvement in detection range. Full-wave simulations and measurements confirm effective non-contact, privacy-preserving short-range human-motion detection across five 28-degree sectors, providing 140-degree total angular coverage. Fall-detection experiments further validate reliable wide-angle performance and continuous spatial tracking. The proposed system offers a compact, low-cost, and scalable platform for millimeter-wave sensing in ambient healthcare and smart-environment applications.

We study language-in-the-loop control for multi-drone systems that execute evolving, high-level missions while retaining formal robustness guarantees at the physical layer. We propose a three-layer architecture in which (i) a human operator issues natural-language instructions, (ii) an LLM-based supervisor periodically interprets, verifies, and corrects the commanded task in the context of the latest state and target estimates, and (iii) a distributed inner-loop controller tracks the resulting reference using only local relative information. We derive a theoretical guarantee that characterizes tracking performance under bounded disturbances and piecewise-smooth references with discrete jumps induced by LLM updates. Overall, our results illustrate how centralized language-based task reasoning can be combined with distributed feedback control to achieve complex behaviors with provable robustness and stability.

We present a falsification framework that integrates learned surrogate dynamics with optimal control to efficiently generate counterexamples for cyber-physical systems specified in signal temporal logic (STL). The unknown system dynamics are identified using neural ODEs, while known a-priori structure is embedded directly into the model, reducing data requirements. The learned neural ODE is converted into an analytical form via symbolic regression, enabling fast and interpretable trajectory optimization. Falsification is cast as minimizing STL robustness over input trajectories; negative robustness yields candidate counterexamples, which are validated on the original system. Spurious traces are iteratively used to refine the surrogate, while true counterexamples are returned as final results. Experiments on ARCH-COMP 2024 benchmarks show that this method requires orders of magnitude fewer experiments of the system under test than optimization-based approaches that do not model system dynamics.

Autoregressive (AR) models excel at generating temporally coherent audio by producing tokens sequentially, yet they often falter in faithfully following complex textual prompts, especially those describing complex sound events. We uncover a surprising capability in AR audio generators: their early prefix tokens implicitly encode global semantic attributes of the final output, such as event count and sound-object category, revealing a form of implicit planning. Building on this insight, we propose Plan-Critic, a lightweight auxiliary model trained with a Generalized Advantage Estimation (GAE)-inspired objective to predict final instruction-following quality from partial generations. At inference time, Plan-Critic enables guided exploration: it evaluates candidate prefixes early, prunes low-fidelity trajectories, and reallocates computation to high-potential planning seeds. Our Plan-Critic-guided sampling achieves up to a 10-point improvement in CLAP score over the AR baseline-establishing a new state of the art in AR text-to-audio generation-while maintaining computational parity with standard best-of-N decoding. This work bridges the gap between causal generation and global semantic alignment, demonstrating that even strictly autoregressive models can plan ahead.

Connectivity in connected and autonomous vehicles (CAVs) introduces vulnerability to cyber threats such as false data injection (FDI) attacks, which can compromise system reliability and safety. To ensure resilience, this paper proposes a control framework combining a nonlinear controller with an interval observer for robust state estimation under measurement noise. The observer bounds leader's states, while a neural network-based estimator estimates the unknown FDI attacks in real time. These estimates are then used to mitigate FDI attack effects maintaining safe inter-vehicle spacing. The proposed approach leverages an idea of interval observer-based estimation and merges model-based and learning-based methods to achieve accurate estimations and real-time performance. MATLAB/Simulink results confirm resilient tracking, precise FDI attack estimation, and robustness to noise, demonstrating potential for real-world CACC applications under cyberattacks, disturbance, and bounded measurement noise.

We propose a novel framework that computes the corrective control efforts to ensure joint safety in multi-agent dynamical systems. This framework efficiently distributes the required corrective effort without revealing individual agents' private preferences. Our framework integrates high-order control barrier functions (HOCBFs), which enforce safety constraints with formal guarantees of safety for complex dynamical systems, with a privacy-preserving resource allocation mechanism based on the progressive second price (PSP) auction. When a joint safety constraint is violated, agents iteratively bid on new corrective efforts via 'avoidance credits' rather than explicitly solving for feasible corrective efforts that remove the safety violation. The resulting correction, determined via a second price payment rule, coincides with the socially optimal safe distribution of corrective actions. Critically, the bidding process achieves this optimal allocation efficiently and without revealing private preferences of individual agents. We demonstrate this method through multi-robot hardware experiments on the Robotarium platform.

Intelligent assistive technologies are increasingly recognized as critical daily-use enablers for people with disabilities and age-related functional decline. Longitudinal studies, curation of quality datasets, live monitoring in activities of daily living, and intelligent intervention devices, share the largely unsolved need in reliable high-throughput multimodal sensing and processing. Streaming large heterogeneous data from distributed sensors, historically closed-source environments, and limited prior works on realtime closed-loop AI methodologies, inhibit such applications. To accelerate the emergence of clinical deployments, we deliver HERMES - an open-source high-performance Python framework for continuous multimodal sensing and AI processing at the edge. It enables synchronized data collection, and realtime streaming inference with user PyTorch models, on commodity computing devices. HERMES is applicable to fixed-lab and free-living environments, of distributed commercial and custom sensors. It is the first work to offer a holistic methodology that bridges cross-disciplinary gaps in real-world implementation strategies, and guides downstream AI model development. Its application on the closed-loop intelligent prosthesis use case illustrates the process of suitable AI model development from the generated constraints and trade-offs. Validation on the use case, with 4 synchronized hosts cooperatively capturing 18 wearable and off-body modalities, demonstrates performance and relevance of HERMES to the trajectory of the intelligent healthcare domain.

Parameter-efficient fine-tuning (PEFT) is a scalable approach for adapting large speech foundation models to new domains. While methods such as LoRA and its state-of-the-art variants reduce adaptation costs, they typically allocate parameters uniformly across model subspaces, which limits their efficiency and scalability in speech applications. Building on our prior work, this paper introduces SSVD-Outer (SSVD-O), an extension of the structured SVD-guided (SSVD) fine-tuning method. SSVD-O combines input acoustic feature space-associated inner transformations with output semantic feature space-associated outer transformations to enable scalable and balanced adaptation. We conduct the first systematic analysis of parameter budget allocation across model subspaces in PEFT for automatic speech recognition (ASR), and investigate the trade-off between learning and forgetting under constrained resources. SSVD-O is benchmarked against LoRA, DoRA, PiSSA, and SSVD on domain-shifted ASR tasks, including child speech and regional accents, across model scales from 0.1B to 2B within the ESPnet framework. Experimental results show that SSVD-O consistently narrows the performance gap to full fine-tuning while improving generalization and mitigating catastrophic forgetting.

Contrastive language-audio pretraining (CLAP) has achieved notable success in learning semantically rich audio representations and is widely adopted for various audio-related tasks. However, current CLAP models face several key limitations. First, they are typically trained on relatively small datasets, often comprising a few million audio samples. Second, existing CLAP models are restricted to short and fixed duration, which constrains their usage in real-world scenarios with variable-duration audio. Third, the standard contrastive training objective operates on global representations, which may hinder the learning of dense, fine-grained audio features. To address these challenges, we introduce Scalable Language-Audio Pretraining (SLAP), which scales language-audio pretraining to 109 million audio-text pairs with variable audio durations and incorporates multiple training objectives. SLAP unifies contrastive loss with additional self-supervised and captioning losses in a single-stage training, facilitating the learning of richer dense audio representations. The proposed SLAP model achieves new state-of-the-art performance on audio-text retrieval and zero-shot audio classification tasks, demonstrating its effectiveness across diverse benchmarks.

The ambiguity of human emotions poses several challenges for machine learning models, as they often overlap and lack clear delineating boundaries. Contrastive language-audio pretraining (CLAP) has emerged as a key technique for generalisable emotion recognition. However, as conventional CLAP enforces a strict one-to-one alignment between paired audio-text samples, it overlooks intra-modal similarity and treats all non-matching pairs as equally negative. This conflicts with the fuzzy boundaries between different emotions. To address this limitation, we propose SmoothCLAP, which introduces softened targets derived from intra-modal similarity and paralinguistic features. By combining these softened targets with conventional contrastive supervision, SmoothCLAP learns embeddings that respect graded emotional relationships, while retaining the same inference pipeline as CLAP. Experiments on eight affective computing tasks across English and German demonstrate that SmoothCLAP is consistently achieving superior performance. Our results highlight that leveraging soft supervision is a promising strategy for building emotion-aware audio-text models.

This paper presents an automated measurement methodology for angular received-power characterization of embedded millimeter-wave transmitters using geometry-calibrated spatial sampling. Characterization of integrated mmWave transmitters remains challenging due to limited angular coverage and alignment variability in conventional probe-station techniques, as well as the impracticality of anechoic-chamber testing for platform-mounted active modules. To address these challenges, we introduce RAPTAR, an autonomous measurement system for angular received-power acquisition under realistic installation constraints. A collaborative robot executes geometry-calibrated, collision-aware hemispherical trajectories while carrying a calibrated receive probe, enabling controlled and repeatable spatial positioning around a fixed device under test. A spectrum-analyzer-based receiver chain acquires amplitude-only received power as a function of angle and distance following quasi-static pose stabilization. The proposed framework enables repeatable angular received-power mapping and power-domain comparison against idealized free-space references derived from full-wave simulation. Experimental results for a 60-GHz radar module demonstrate a mean absolute received-power error below 2 dB relative to simulation-derived references and a 36.5 % reduction in error compared to manual probe-station measurements, attributed primarily to reduced alignment variability and consistent spatial sampling. The proposed method eliminates the need for coherent field measurements and near-field transformations, enabling practical power-domain characterization of embedded mmWave modules. It is well suited for angular validation in real-world platforms where conventional anechoic measurements are impractical.

We propose PISE, a physics-informed deep ghost imaging framework for low-bandwidth edge perception. By combining adjoint operator initialization with semantic guidance, PISE improves classification accuracy by 2.57% and reduces variance by 9x at 5% sampling.

The purpose of this paper is to illustrate, in an experimental facility consisting of a simple pendular device, the performance of a sliding mode adaptive position-feedback tracking controller of mechanical systems with friction reported in the literature. To put this experimental evidence in perspective, we compare the performance of the sliding mode scheme with the one obtained by an adaptive controller designed following the well-known immersion and invariance technique.

Modulo-Imaging (MI) offers a promising alternative for expanding the dynamic range of images by resetting the signal intensity when it reaches the saturation level. Subsequently, high-dynamic range (HDR) modulo imaging requires a recovery process to obtain the HDR image. MI is a non-convex and ill-posed problem where recent recovery networks suffer in high-noise scenarios. In this work, we formulate the HDR reconstruction task as an optimization problem that incorporates a deep prior and subsequently unrolls it into an optimization-inspired deep neural network. The network employs a lightweight convolutional denoiser for fast inference with minimal computational overhead, effectively recovering intensity values while mitigating noise. Moreover, we introduce the Scaling Equivariance term that facilitates self-supervised fine-tuning, thereby enabling the model to adapt to new modulo images that fall outside the original training distribution. Extensive evaluations demonstrate the superiority of our method compared to state-of-the-art recovery algorithms in terms of performance and quality.

Audio large language models (LLMs) enable unified speech understanding and generation, but adapting them to linguistically complex and dialect-rich settings such as Arabic-English remains challenging. We present a controlled study of multi-task instruction tuning for an Arabic-centric audio LLM across generative tasks, including automatic speech recognition (ASR) and speech and text summarization, as well as discriminative tasks, including dialect identification (DID) and speech emotion recognition (SER), in a resource-constrained setting. To support end-to-end Arabic speech summarization, we introduce AraMega-SSum, the first Arabic speech summarization dataset designed for training and benchmarking Arabic-centric audio LLMs. We compare four training strategies: (i) Uniform Mixing (UM), (ii) Task-Progressive Curriculum (TPC), (iii) Aligner-Based Diverse Sampling (ADS) for training-time batch construction, and (iv) a two-stage TPC->ADS strategy. Our results reveal a clear efficiency-robustness trade-off. TPC achieves the strongest performance on generative tasks, including ASR and summarization. ADS improves paralinguistic tasks but reduces generative stability when used alone. The two-stage TPC->ADS strategy provides the best overall balance, achieving the strongest DID and SER performance while outperforming large proprietary models such as Gemini-2.5-Pro on discriminative tasks. We will publicly release AraMega-SSum together with all experimental resources to support future research in Arabic speech understanding.

Online blind source separation is essential for both speech communication and human-machine interaction. Among existing approaches, overdetermined independent vector analysis (OverIVA) delivers strong performance by exploiting the statistical independence of source signals and the orthogonality between source and noise subspaces. However, when applied to large microphone arrays, the number of parameters grows rapidly, which can degrade online estimation accuracy. To overcome this challenge, we propose decomposing each long separation filter into a bilinear form of two shorter filters, thereby reducing the number of parameters. Because the two filters are closely coupled, we design an alternating iterative projection algorithm to update them in turn. Simulation results show that, with far fewer parameters, the proposed method achieves improved performance and robustness.

This paper analyzes the impact of spatially correlated additive noise on the minimum mean-square error (MMSE) estimation of multiple-input multiple-output (MIMO) channels from one-bit quantized observations. Although additive noise can be correlated in practical scenarios, e.g., due to jamming, clutter, or other external disturbances, the effect of such correlation on the MMSE channel estimator in this setting remains unexplored in prior work. Against this backdrop, we derive a novel analytical expression for the general MIMO MMSE channel estimator, which is inherently nonlinear in one-bit observations, and accommodates arbitrary channel and noise correlation structures. To further characterize the impact of noise correlation, we subsequently specialize the general MMSE expression to certain tractable multi antenna configurations in which both the channel and the noise assume single-parameter constant correlation structures. Our analyses reveal nontrivial, noise-correlation-induced scenarios in which the estimator remains linear despite non-zero channel and noise correlation parameters. Moreover, the results indicate that, at low-to-medium signal-to-noise ratio, noise correlation improves the MMSE performance when channels are uncorrelated, but degrades performance when channels are strongly correlated.

Neural codec language models achieve impressive zero-shot Text-to-Speech (TTS) by fully imitating the acoustic characteristics of a short speech prompt, including timbre, prosody, and paralinguistic information. However, such holistic imitation limits their ability to isolate and control individual attributes. In this paper, we present a unified codec language model SpeechEdit that extends zero-shot TTS with a selective control mechanism. By default, SpeechEdit reproduces the complete acoustic profile inferred from the speech prompt, but it selectively overrides only the attributes specified by explicit control instructions. To enable controllable modeling, SpeechEdit is trained on our newly constructed LibriEdit dataset, which provides delta (difference-aware) training pairs derived from LibriHeavy. Experimental results show that our approach maintains naturalness and robustness while offering flexible and localized control over desired attributes. Audio samples are available at https://speech-editing.github.io/speech-editing/.

Audio-visual speech recognition (AVSR) typically improves recognition accuracy in noisy environments by integrating noise-immune visual cues with audio signals. Nevertheless, high-noise audio inputs are prone to introducing adverse interference into the feature fusion process. To mitigate this, recent AVSR methods often adopt mask-based strategies to filter audio noise during feature interaction and fusion, yet such methods risk discarding semantically relevant information alongside noise. In this work, we propose an end-to-end noise-robust AVSR framework coupled with speech enhancement, eliminating the need for explicit noise mask generation. This framework leverages a Conformer-based bottleneck fusion module to implicitly refine noisy audio features with video assistance. By reducing modality redundancy and enhancing inter-modal interactions, our method preserves speech semantic integrity to achieve robust recognition performance. Experimental evaluations on the public LRS3 benchmark suggest that our method outperforms prior advanced mask-based baselines under noisy conditions.

Reliable flow measurements are essential in many industries, but current instruments often fail to accurately estimate multiphase flows, which are frequently encountered in real-world operations. Combining machine learning (ML) algorithms with accurate single-phase flowmeters has therefore received extensive research attention in recent years. The Coriolis mass flowmeter is a widely used single-phase meter that provides direct mass flow measurements, which ML models can be trained to correct, thereby reducing measurement errors in multiphase conditions. This paper demonstrates that preserving temporal information significantly improves model performance in such scenarios. We compare a multilayer perceptron, a windowed multilayer perceptron, and a convolutional neural network (CNN) on three-phase air-water-oil flow data from 342 experiments. Whereas prior work typically compresses each experiment into a single averaged sample, we instead compute short-time averages from within each experiment and train models that preserve temporal information at several downsampling intervals. The CNN performed best at 0.25 Hz with approximately 95 % of relative errors below 13 %, a normalized root mean squared error of 0.03, and a mean absolute percentage error of approximately 4.3 %, clearly outperforming the best single-averaged model and demonstrating that short-time averaging within individual experiments is preferable. Results are consistent across multiple data splits and random seeds, demonstrating robustness.

This paper studies a reconfigurable intelligent surface (RIS)-enhanced decoupled symbiotic radio (SR) system in which a primary transmitter delivers common data to multiple primary receivers (PRs), while a RIS-based backscatter device sends secondary data to a backscatter receiver (BRx). Unlike conventional SR, the BRx performs energy detection and never decodes the primary signal, thereby removing ambiguity and preventing exposure of the primary payload to unintended receivers. In this paper, we formulate the problem as the minimization of the transmit power subject to a common broadcast rate constraint across all PRs and a bit error rate (BER) constraint at the BRx. The problem is nonconvex due to the unit-modulus RIS constraint and coupled quadratic forms. Leveraging a rate-balanced reformulation and a monotonic BER ratio characterization, we develop a low-complexity penalty-based block coordinate descent algorithm with closed-form updates. Numerical results show fast convergence of the proposed algorithm and reduced power consumption of the considered RIS-enhanced information-decoupled SR system over conventional SR baselines.