Papers

Accurate segmentation of infant brain MRI is critical for studying early neurodevelopment and diagnosing neurological disorders. Yet, it remains a fundamental challenge due to continuously evolving anatomy of the subjects, motion artifacts, and the scarcity of high-quality labeled data. In this work, we present LODi, a novel framework that utilizes prior knowledge from an adult brain MRI segmentation model to enhance the segmentation performance of infant scans. Given the abundance of publicly available adult brain MRI data, we pre-train a segmentation model on a large adult dataset as a starting point. Through transfer learning and domain adaptation strategies, we progressively adapt the model to the 0-2 year-old population, enabling it to account for the anatomical and imaging variability typical of infant scans. The adaptation of the adult model is carried out using weakly supervised learning on infant brain scans, leveraging silver-standard ground truth labels obtained with FreeSurfer. By introducing a novel training strategy that integrates hierarchical feature refinement and multi-level consistency constraints, our method enables fast, accurate, age-adaptive segmentation, while mitigating scanner and site-specific biases. Extensive experiments on both internal and external datasets demonstrate the superiority of our approach over traditional supervised learning and domain-specific models. Our findings highlight the advantage of leveraging adult brain priors as a foundation for age-flexible neuroimaging analysis, paving the way for more reliable and generalizable brain MRI segmentation across the lifespan.

The fundamental role of power converters is to efficiently manage and control the flow of electrical energy, ensuring compatibility between power sources and loads. All these applications of power converters need the design of an appropriate control law. Control of power converters is a challenging problem due to the presence of switching devices which are difficult to handle using traditional control approaches. The objective of this paper is to investigate the use of data-driven techniques, in particular the Virtual References Feedback Tuning (VRFT) method, in the context of power converters feedback control. This study considers a buck \pauline{mode} power converter circuit provided by the OwnTech foundation.

Animals often forage via Levy walks stochastic trajectories with heavy tailed step lengths optimized for sparse resource environments. We show that human visual gaze follows similar dynamics when scanning images. While traditional models emphasize image based saliency, the underlying spatiotemporal statistics of eye movements remain underexplored. Understanding these dynamics has broad applications in attention modeling and vision-based interfaces. In this study, we conducted a large scale human subject experiment involving 40 participants viewing 50 diverse images under unconstrained conditions, recording over 4 million gaze points using a high speed eye tracker. Analysis of these data shows that the gaze trajectory of the human eye also follows a Levy walk akin to animal foraging. This suggests that the human eye forages for visual information in an optimally efficient manner. Further, we trained a convolutional neural network (CNN) to predict fixation heatmaps from image input alone. The model accurately reproduced salient fixation regions across novel images, demonstrating that key components of gaze behavior are learnable from visual structure alone. Our findings present new evidence that human visual exploration obeys statistical laws analogous to natural foraging and open avenues for modeling gaze through generative and predictive frameworks.

Predictive Stator Current Control (PSCC) has been proposed for control of multi-phase drives. The flexibility offered by the use of a Cost Function has been used to deal with the increased number of phases. However, tuning of the Weighting Factors constitutes a problem. Intensive trial and error tests are usual in this context. Existing on-line selection methods, on the other hand, require large amounts of data and/or complex optimization procedures. The proposal of this paper is a closed-loop scheme that links Weighting Factors to performance indicators. In this way, optimal Weighting Factors are determined for each operating point. Also, changes in reference values for performance indicators are easily tackled. Unlike previous methods, the proposal carries very little computational burden. A case study is developed for a five-phase induction motor and assessed with real experimentation on a laboratory set-up.

Electric Vertical Take-Off and Landing (eVTOL) aircraft are expected to be quieter and more cost-effective than helicopters, offering major economic and social benefits through improved connectivity. Their adoption will require new ground infrastructure and airspace redesign, introducing risks involving multiple stakeholders (Regulators, eVTOL operators, Air navigation service providers, Vertiport operators, OEMs, Pilots, etc.). To assess these risks for the UK airspace, systems-thinking based System Theoretic Process Analysis (STPA) was conducted. To manage the large number of Unsafe Control Actions (UCAs) and requirements generated due to the complexity of the analysis, a novel extension to STPA for the prioritization of results was applied. 317 UCAs were identified in total out of which 110 high-priority UCAs were analyzed (Step-4), resulting in 377 causal factors and 432 requirements. These were prioritized to produce a targeted list of 124 distinct high-priority requirements, 56 of which were identified as gaps in existing aviation regulations, policies, or procedures.. These highlight opportunities for regulatory updates in areas such as organizational performance, certification processes, training, collision avoidance, energy management, and automation. The findings provide regulators with safety considerations that could shape new or updated regulations, compliance methods, and guidance materials for the safe deployment of eVTOLs.

The field of Finite State Model Predictive Control for multiphase drives has produced many contributions. Many variants of FSMPC exist, each aiming at some aspect such as complexity of the cost function, switching frequency, etc. Despite past efforts to compare different techniques, the field is still out of consensus regarding the relative merits of each one. This paper presents a new method to compare FSMPC variants. The method is based on analyzing the modulation, implicit or explicit, used by each variant. In the paper the method is used to compare single-vector state-of-the-art FSMPC with a multi-vector variant designed to cancel xy currents and simplify the cost function. The results show the strengths and weaknesses of each technique. Also, it is found that the trade-offs between figures, previously thought to concern just individual regimes, extend to the whole operating space and also can be pinpoint to each FSMPC variant. Finally, it is shown that the flexibility of the single-vector approach and its better DC-link usage makes it, arguably, superior over the multi-vector variant.

Voice Conversion (VC) aims to modify a speaker's timbre while preserving linguistic content. While recent VC models achieve strong performance, most struggle in real-time streaming scenarios due to high latency, dependence on ASR modules, or complex speaker disentanglement, which often results in timbre leakage or degraded naturalness. We present SynthVC, a streaming end-to-end VC framework that directly learns speaker timbre transformation from synthetic parallel data generated by a pre-trained zero-shot VC model. This design eliminates the need for explicit content-speaker separation or recognition modules. Built upon a neural audio codec architecture, SynthVC supports low-latency streaming inference with high output fidelity. Experimental results show that SynthVC outperforms baseline streaming VC systems in both naturalness and speaker similarity, achieving an end-to-end latency of just 77.1 ms.

Maintaining high energy efficiency (EE) in wireless networks is crucial, particularly with the adoption of massive MIMO technology. This work introduces a resource allocation framework that jointly optimizes transmit power assigned to each user and the number of active antennas, while explicitly accounting for a nonlinear Power Amplifier (PA). We consider a downlink MU-MIMO-OFDM transmission with zero forcing (ZF) precoding, Rayleigh fading channels, and soft-limiter PAs, with both ideal and realistic PA architectures. In contrast to existing formulations, our optimization framework avoids imposing an explicit transmit power constraint, since the nonlinear distortion inherently limits the feasible operating region. To solve the resulting non-convex problem, an alternating optimization approach is adopted that, by exploiting properties of the EE function, guarantees convergence to a stationary point. Extensive simulations demonstrate consistent performance gains over distortion-neglecting and power-only optimized baselines. In a scenario of a 5 km radius cell serving 60 randomly distributed users, the median EE gains over the distortion-neglecting allocation reach 40% for ideal PAs and 20% for Class B PAs, confirming high impact of the proposed solution.

Upon choosing microphones for automotive hands-free communication or Automatic Speech Recognition (ASR) applications, OEMs typically specify wideband, super wideband or even fullband requirements following established standard recommendations (e.g., ITU-P.1110, ITU-P.1120). In practice, it is often challenging to achieve the preferred bandwidth for an automotive microphone when considering limitations and constraints on microphone placement inside the cabin, and the automotive grade environmental robustness requirements. On the other hand, there seems to be no consensus or sufficient data on the effect of each microphone characteristic on the actual performance. As an attempt to answer this question, we used noise signals recorded in real vehicles and under various driving conditions to experimentally study the relationship between the microphones' characteristics and the final audio quality of speech communication and performance of ASR engines. We focus on how variations in microphone bandwidth and amplitude frequency response shapes affect the perceptual speech quality. The speech quality results are compared by using ETSI TS 103 281 metrics (S-MOS, N-MOS, G-MOS) and ancillary metrics such as SNR. The ASR results are evaluated with standard metrics such as Word Error Rate (WER). Findings from this study provide knowledge in the understanding of what microphone frequency response characteristics are more relevant for audio quality and choice of proper microphone specifications, particularly for automotive applications.

Although multiple works have proposed energy-efficient resource allocation schemes for Massive Multiple-Input Multiple-Output (M-MIMO) system, most approaches overlook the potential of optimizing Power Amplifier (PA) transmission power while accounting for non-linear distortion effects. Furthermore, most M-MIMO studies assume narrow-band transmission, neglecting subcarrier intermodulations at the non-linear PA for an Orthogonal Frequency Division Multiplexing (OFDM) system. Therefore, this work investigates the energy-efficient power allocation for a single-user equipment (UE) M-MIMO downlink (DL) system employing OFDM with nonlinear PAs. Unlike prior works, we model wide-band transmission using a soft-limiter PA model and derive a closed-form expression for the signal-to-distortion-and-noise ratio (SNDR) under Rayleigh fading and Maximal Ratio Transmission (MRT) precoding. Next, the Energy Efficiency (EE) function is defined considering two PA architectures and a distorted OFDM signal. We then propose a low complexity root-finding algorithm to maximize EE by transmit power adjustment. Simulation results demonstrate significant EE gains over a fixed PA back-off baseline, with over $100\%$ improvement under both low and high path loss. Our findings reveal how the optimal operating point depends on the antenna count, the PA model, and the propagation conditions.

Zero-resource word segmentation and clustering systems aim to tokenise speech into word-like units without access to text labels. Despite progress, the induced lexicons are still far from perfect. In an idealised setting with gold word boundaries, we ask whether performance is limited by the representation of word segments, or by the clustering methods that group them into word-like types. We combine a range of self-supervised speech features (continuous/discrete, frame/word-level) with different clustering methods (K-means, hierarchical, graph-based) on English and Mandarin data. The best system uses graph clustering with dynamic time warping on continuous features. Faster alternatives use graph clustering with cosine distance on averaged continuous features or edit distance on discrete unit sequences. Through controlled experiments that isolate either the representations or the clustering method, we demonstrate that representation variability across segments of the same word type – rather than clustering – is the primary factor limiting performance.

Physical unclonable functions (PUFs) involve challenging practical applications of error-correcting codes (ECCs), requiring extremely low failure rates on the order of $10^{-6}$ and below despite raw input bit error rates as high as 22%. These requirements call for an efficient ultra-low rate code design. In this work, we propose a novel coding scheme tailored for PUFs based on Polar codes and a low-complexity version of automorphism ensemble decoding (AED). Notably, our serial AED scheme reuses a single successive cancellation (SC) decoder across multiple decoding attempts. By introducing cascaded and recursive interleavers, we efficiently scale the number of AED candidates without requiring expensive large multiplexers. An aggressive quantization strategy of only 3 bits per message further reduces the area requirements of the underlying SC decoder. The resulting coding scheme achieves the same block error rate of $10^{-6}$ as our baseline based on Bose-Ray-Chaudhuri-Hocquenghem (BCH) codes while requiring 1.75x fewer codeword bits to encode the same K = 312 payload bits. This reduction translates directly into 1.75x less helper data storage and, consequently, a smaller overall chip area.

In this paper, we consider the problem of estimating the delay-Doppler (DD) domain input-output (I/O) relation in Zak-OTFS modulation, which is needed for signal detection. Two approaches, namely, model-dependent and model-free approaches, can be employed for this purpose. The model-dependent approach requires explicit estimation of the physical channel parameters (path delays, Dopplers, and gains) to obtain the I/O relation. Such an explicit estimation is not required in the model-free approach, where the I/O relation can be estimated by reading off the samples in the fundamental DD period of the received pilot frame. Model-free approach has the advantage of acquiring fractional DD channels with simplicity. However, the read-off in the model-free approach provides an estimate of the effective channel only over a limited region in the DD plane but it does not provide an estimate for the region outside, and this can affect the estimation performance depending on the pulse shaping characteristics of the DD pulse shaping filter used. A poorly localized DD pulse shape leads to an increased degradation in performance. Motivated by this, in this paper, we propose a novel, yet simple, I/O relation estimation scheme that alleviates the above issue in the model-free approach. We achieve this by obtaining a coarse estimate of the effective channel outside the model-free estimation region using a novel model-dependent scheme and using this estimate along with the model-free estimate to obtain an improved estimate of the overall I/O relation. We devise the proposed estimation scheme for both exclusive and embedded pilot frames. Our simulation results using Vehicular-A, TDL-A and TDL-C channel models with fractional DDs show that the proposed hybrid estimation approach achieves superior performance compared to the pure model-free approach.

Transient measurements, captured by the timeresolved systems, are widely employed in photon-efficient reconstruction tasks, including line-of-sight (LOS) and non-line-of-sight (NLOS) imaging. However, challenges persist in their 3D reconstruction due to the low quantum efficiency of sensors and the high noise levels, particularly for long-range or complex scenes. To boost the 3D reconstruction performance in photon-efficient imaging, we propose a generic Time-Resolved Transformer (TRT) architecture. Different from existing transformers designed for high-dimensional data, TRT has two elaborate attention designs tailored for the spatio-temporal transient measurements. Specifically, the spatio-temporal self-attention encoders explore both local and global correlations within transient data by splitting or downsampling input features into different scales. Then, the spatio-temporal cross attention decoders integrate the local and global features in the token space, resulting in deep features with high representation capabilities. Building on TRT, we develop two task-specific embodiments: TRT-LOS for LOS imaging and TRT-NLOS for NLOS imaging. Extensive experiments demonstrate that both embodiments significantly outperform existing methods on synthetic data and real-world data captured by different imaging systems. In addition, we contribute a large-scale, high-resolution synthetic LOS dataset with various noise levels and capture a set of real-world NLOS measurements using a custom-built imaging system, enhancing the data diversity in this field. Code and datasets are available at https://github.com/Depth2World/TRT.

Graph learning aims to infer a network structure directly from observed data, enabling the analysis of complex dependencies in irregular domains. Traditional methods focus on scalar signals at each node, ignoring dependencies along additional dimensions such as time, configurations of the observation device, or populations. In this work, we propose a graph signal processing framework for learning graphs from two-dimensional signals, modeled as matrix graph signals generated by joint filtering along both dimensions. This formulation leverages the concept of graph stationarity across the two dimensions and leverages product graph representations to capture structured dependencies. Based on this model, we design an optimization problem that can be solved efficiently and provably recovers the optimal underlying Kronecker/Cartesian/strong product graphs. Experiments on synthetic data demonstrate that our approach achieves higher estimation accuracy and reduced computational cost compared to existing methods.

A sliding-mode-based adaptive boundary control law is proposed for a class of uncertain thermal reaction-diffusion processes subject to matched disturbances. The disturbances are assumed to be bounded, but the corresponding bounds are unknown, thus motivating the use of adaptive control strategies. A boundary control law comprising a proportional and discontinuous term is proposed, wherein the magnitude of the discontinuous relay term is adjusted via a gradient-based adaptation algorithm. Depending on how the adaptation algorithm is parameterized, the adaptive gain can be either a nondecreasing function of time (monodirectional adaptation) or it can both increase and decrease (bidirectional adaptation). The convergence and stability properties of these two solutions are investigated by Lyapunov analyses, and two distinct stability results are derived, namely, asymptotic stability for the monodirectional adaptation and globally uniformly ultimately bounded solutions for the bidirectional adaptation. The proposed algorithms are then specified to address the control problem of stabilizing a desired temperature profile in a metal beam equipped with thermoelectric boundary actuators. Experiments are conducted to investigate the real-world performance of the proposed sliding-mode-based adaptive control, with a particular focus on comparing the monodirectional and bidirectional adaptation laws.

The objective of Audio Augmented Reality (AAR) applications are to seamlessly integrate virtual sound sources within a real environment. It is critical for these applications that virtual sources are localised precisely at the intended position, and that the acoustic environments are accurately matched. One effective method for spatialising sound on headphones is through Head-Related Transfer Functions (HRTFs). These characterise how the physical features of a listener modify sound waves before they reach the eardrum. This study examines the influence of using individualised HRTFs on the localisation and the perceived realism of virtual sound sources associated with a real visual object. Participants were tasked with localising virtual and real speech sources presented via headphones and through a spherical loudspeaker array, respectively. The assessment focussed on perceived realism and sources location. All sources were associated with one of thirty real visual sources (loudspeakers) arranged in a semi-anechoic room. Various sound source renderings were compared, including single loudspeaker rendering and binaural rendering with individualised or non-individualised HRTFs. Additionally, the impact of head movements was explored: ten participants completed the same task with and without the possibility to move their head. The results showed that using individual HRTFs improved perceived realism but not localisation performance in the static scenario. Surprisingly, the opposite was observed when head movements were possible and encouraged.

High Electron Mobility Transistors (HEMTs) are most suitable for harsh environments as they operate reliably under extreme conditions such as high voltages, high temperatures, radiation exposure and corrosive atmospheres. In this article, gate field-plated engineering Al0.295GaN/GaN HEMT is proposed for achieving high breakdown voltage to reliably operate in harsh environments. The Al0.295GaN/GaN heterointerface results in a 2DEG (two-dimensional electron gas) density of the order of 1013 cm-2 obtained from the self-consistent solution of Schrödinger and Poisson equations. The device has undergone DC and breakdown simulations which result in threshold voltage of -5.5 V, drain saturation current of 3000 mA, and breakdown voltage of 1 kV. The HEMT also shows excellent RF characteristics which include cut-off frequency (ft) of 28 GHz and maximum frequency of oscillation (fmax) of 38 GHz. The proposed gate field-plated HEMT is stable up to 40 GHz and suitable for high-voltage and high-power RF operation during harsh environment applications.

One of the key challenges in many research fields is uncovering how different interconnected systems interact within complex networks, typically represented as multi-layer networks. Capturing the intra- and cross-layer interactions among different domains for analysis and processing calls for topological algebraic descriptors capable of localizing the homologies of different domains, at different scales, according to the learning task. Our first contribution in this paper is to introduce the Cell MultiComplexes (CMCs), which are novel topological spaces that enable the representation of higher-order interactions among interconnected cell complexes. We introduce cross-Laplacian operators as powerful algebraic descriptors of CMC spaces able to capture different topological invariants, whether global or local, at different resolutions. Using the eigenvectors of these operators as bases for the signal representation, we develop topological signal processing tools for signals defined over CMCs. Then, we focus on the signal spectral representation and on the filtering of noisy flows observed over the cross-edges between different layers of CMCs. We show that a local signal representation based on cross-Laplacians yields a better sparsity/accuracy trade-off compared to monocomplex representations, which provide overcomplete representation of local signals. Finally, we illustrate a topology learning strategy designed to infer second-order cross-cells between layers, with applications to brain networks for encoding inter-module connectivity patterns.

Numerous types of antennas have been employed for microwave imaging of stratified media for ground penetrating radar (GPR), through-the-wall-radar imaging (TWRI), etc. This letter aims to investigate the impact of the different antennas with their characteristics on the image reconstruction of those media. Hence, three types of antennas, including horn antennas, open waveguide and Vivaldi antennas, are chosen as almost directional antennas, operating at X-band 8-12 GHz. The antenna's far-field and near-field characteristics are analyzed. A diffraction tomography (DT)-based algorithm is used to reconstruct the target location within the stratified media using monostatic and multistatic data. It is observed that the more directional antennas provide a better-reconstructed image with less shadowing image of the stratified media.

The fundamental sensing limit of pinching-antenna systems (PASS) is studied from a Bayesian Cramér-Rao bound (BCRB) perspective. Compared to conventional CRB, BCRB is independent of the exact values of sensing parameters and is not restricted by the unbiasedness of the estimator, thus offering a practical and comprehensive lower bound for evaluating sensing performance. A system where multiple targets transmit uplink pilots to a single-waveguide PASS under a time-division multiple access (TDMA) scheme is analyzed. For the single-target scenario, our analysis reveals a unique mismatch between the sensing centroid (i.e., the optimal PA position) and the distribution centroid (i.e., the center of the target's prior distribution), underscoring the necessity of dynamic PA repositioning. For the multi-target scenario, two target scheduling protocols are proposed: 1) pinch switching (PS), which performs separate pinching beamforming for each time slot, and 2) pinch multiplexing (PM), which applies a single beamforming configuration across all slots. Based on these protocols, both the total power minimization problem under a BCRB threshold and the min-max BCRB problem under a total power constraint are formulated. By leveraging Karush-Kuhn-Tucker (KKT) conditions, these problems are equivalently converted into a search over PA positions and solved using an element-wise algorithm. Numerical results show that i)~PASS, endowed with large-scale reconfigurability, can significantly enhance the sensing performance compared with conventional fixed-position arrays, and ii)~PS provides more robust performances than PM at the cost of higher computational complexity.

CTC-based ASR systems face computational and memory bottlenecks in resource-limited environments. Traditional CTC decoders, requiring up to 90% of processing time in systems (e.g., wav2vec2-large on L4 GPUs), face inefficiencies due to exhaustive token-level operations. This paper introduces Frame Level Token Pruning for Connectionist Temporal Classification (FLToP CTC), a novel decoding algorithm that employs frame-level token pruning guided by a relative threshold probability. By dynamically eliminating low-probability tokens per frame, FLToP CTC reduces compute and memory demands while maintaining negligible WER degradation. On LibriSpeech, FLToP CTC achieves a 10.5x runtime speedup and 2.78x memory reduction versus standard CTC decoders. Its simplicity enables seamless integration into CTC decoders across platforms (CPUs, GPUs, etc.). FLToP CTC addresses CTC bottlenecks, offering scalability for resource-limited environments and realtime applications, enhancing speech recognition accessibility and efficiency.

Effectiveness of speech emotion recognition in real-world scenarios is often hindered by noisy environments and variability across datasets. This paper introduces a two-step approach to enhance the robustness and generalization of speech emotion recognition models through improved representation learning. First, our model employs EDRL (Emotion-Disentangled Representation Learning) to extract class-specific discriminative features while preserving shared similarities across emotion categories. Next, MEA (Multiblock Embedding Alignment) refines these representations by projecting them into a joint discriminative latent subspace that maximizes covariance with the original speech input. The learned EDRL-MEA embeddings are subsequently used to train an emotion classifier using clean samples from publicly available datasets, and are evaluated on unseen noisy and cross-corpus speech samples. Improved performance under these challenging conditions demonstrates the effectiveness of the proposed method.

We introduce MMAudioSep, a generative model for video/text-queried sound separation that is founded on a pretrained video-to-audio model. By leveraging knowledge about the relationship between video/text and audio learned through a pretrained audio generative model, we can train the model more efficiently, i.e., the model does not need to be trained from scratch. We evaluate the performance of MMAudioSep by comparing it to existing separation models, including models based on both deterministic and generative approaches, and find it is superior to the baseline models. Furthermore, we demonstrate that even after acquiring functionality for sound separation via fine-tuning, the model retains the ability for original video-to-audio generation. This highlights the potential of foundational sound generation models to be adopted for sound-related downstream tasks. Our code is available at https://github.com/sony/mmaudiosep.

Traditional voice conversion (VC) methods typically attempt to separate speaker identity and linguistic information into distinct representations, which are then combined to reconstruct the audio. However, effectively disentangling these factors remains challenging, often leading to information loss during training. In this paper, we propose a new approach that leverages synthetic speech data generated by a high-quality, pretrained multispeaker text-to-speech (TTS) model. Specifically, synthetic data pairs that share the same linguistic content but differ in speaker identity are used as input-output pairs to train the voice conversion model. This enables the model to learn a direct mapping between source and target voices, effectively capturing speaker-specific characteristics while preserving linguistic content. Additionally, we introduce a flexible training strategy for any-to-any voice conversion that generalizes well to unseen speakers and new languages, enhancing adaptability and performance in zero-shot scenarios. Our experiments show that our proposed method achieves a 16.35% relative reduction in word error rate and a 5.91% improvement in speaker cosine similarity, outperforming several state-of-the-art methods. Voice conversion samples can be accessed at: https://oovc-emnlp-2025.github.io/

The rapid growth of the low-altitude economy has resulted in a significant increase in the number of Low, slow, and small (LLS) unmanned aerial vehicles (UAVs), raising critical challenges for secure airspace management and reliable trajectory planning. To address this, this paper proposes a cooperative radio-frequency (RF) detection and localization framework that leverages existing cellular base stations. The proposed approach features a robust scheme for LSS target identification, integrating a cell averaging-constant false alarm rate (CA-CFAR) detector with a micro-Doppler signature (MDS) based recognition method. Multi-station measurements are fused through a grid-based probabilistic algorithm combined with clustering techniques, effectively mitigating ghost targets and improving localization accuracy in multi-UAV scenarios. Furthermore, the Cramer-Rao lower bound (CRLB) is derived as a performance benchmark and reinforcement learning (RL)-based optimization is employed to balance localization accuracy against station resource usage. Simulations demonstrate that increasing from one to multiple BSs reduces the positioning error to near the CRLB, while practical experiments further verify the framework's effectiveness. Furthermore, our RL-based optimization can find solutions that maintain high accuracy while minimizing resource usage, highlighting its potential as a scalable solution for ensuring airspace safety in the emerging low-altitude economy.

We propose a transfer learning-enabled Transformer framework to simultaneously realize accurate modeling and Raman pump design in C+L-band systems. The RMSE for modeling and peak-to-peak GSNR variation/deviation is within 0.22 dB and 0.86/0.1 dB, respectively.

In this work, we propose a meta-learning-based Koopman modeling and predictive control approach for nonlinear systems with parametric uncertainties. An adaptive deep meta-learning-based modeling approach, called Meta Adaptive Koopman Operator (MAKO), is proposed. Without knowledge of the parametric uncertainty, the proposed MAKO approach can learn a meta-model from a multi-modal dataset and efficiently adapt to new systems with previously unseen parameter settings by using online data. Based on the learned meta Koopman model, a predictive control scheme is developed, and the stability of the closed-loop system is ensured even in the presence of previously unseen parameter settings. Through extensive simulations, our proposed approach demonstrates superior performance in both modeling accuracy and control efficacy as compared to competitive baselines.

This paper introduces a new training strategy to improve speech dereverberation systems in an unsupervised manner using only reverberant speech. Most existing algorithms rely on paired dry/reverberant data, which is difficult to obtain. Our approach uses limited acoustic information, like the reverberation time (RT60), to train a dereverberation system. Experimental results demonstrate that our method achieves more consistent performance across various objective metrics than the state-of-the-art.

Recent progress in diffusion-based Singing Voice Synthesis (SVS) demonstrates strong expressiveness but remains limited by data scarcity and model scalability. We introduce a two-stage pipeline: a compact seed set of human-sung recordings is constructed by pairing fixed melodies with diverse LLM-generated lyrics, and melody-specific models are trained to synthesize over 500 hours of high-quality Chinese singing data. Building on this corpus, we propose DiTSinger, a Diffusion Transformer with RoPE and qk-norm, systematically scaled in depth, width, and resolution for enhanced fidelity. Furthermore, we design an implicit alignment mechanism that obviates phoneme-level duration labels by constraining phoneme-to-acoustic attention within character-level spans, thereby improving robustness under noisy or uncertain alignments. Extensive experiments validate that our approach enables scalable, alignment-free, and high-fidelity SVS.