Subject: Cutting-Edge Advancements in Signal Processing and Communication

Hi Elman,

General Overview

This collection of papers explores diverse facets of signal processing and communication, with a notable emphasis on emerging technologies for 6G and beyond. Several studies delve into integrated sensing and communication (ISAC), highlighting its potential to revolutionize wireless networks. Zhang et al. (2025) propose a multi-perspective observation framework for multi-RIS-assisted ISAC, leveraging symbol-level precoding and space-time adaptive processing to enhance target detection while guaranteeing communication quality. Asaad and Tabassum (2025) introduce a robust design for over-the-air federated edge learning (OTA-FEEL) within ISAC, using sensing to mitigate target echoes during model aggregation. Ma et al. (2025) explore the benefits of movable antenna arrays in ISAC, demonstrating improved communication and sensing performance through antenna position optimization. Morrison Atsu et al. (2025) employ multi-agent reinforcement learning (MARL) to optimize UAV swarm trajectories in bistatic ISAC systems, enhancing sensing estimation via decentralized cooperative strategies and transmission power adaptation.

Beyond ISAC, several papers tackle specific challenges in wireless communication. Han et al. (2025) propose an RSMA-aided precoding method for FDD MIMO systems that reconstructs downlink CSI from uplink signals, aiming to reduce latency in IoT communications by eliminating CSIT feedback. Wang et al. (2025) introduce a multi-carrier faster-than-Nyquist (MC-FTN) signaling scheme for OTFS systems, boosting capacity through non-orthogonal pulses. Chen et al. (2025) investigate optimal antenna spacing in near-field XL-MIMO systems to maximize effective degrees-of-freedom and array gain. Song et al. (2025) propose a beam structured turbo receiver for HF skywave massive MIMO, achieving low-complexity signal detection through beam domain processing. Cheng et al. (2025) introduce SynthSoM, a synthetic multi-modal sensing-communication dataset to facilitate research on Synesthesia of Machines (SoM), a concept further explored by Han et al. (2025) in the context of V2V channel modeling.

Novel signal processing techniques are also prominent. Zheng et al. (2025) present a Siamese neural network for single-snapshot DOA estimation with sparse linear arrays, addressing limited data challenges in dynamic environments. Tadik et al. (2025) introduce OpenGERT, an open-source tool for automated geometry extraction, enabling high-fidelity ray-tracing simulations and sensitivity analyses for urban propagation modeling. Singh et al. (2025) propose a Gaussian integral based Bayesian smoother for nonlinear stochastic state-space models, improving accuracy over existing Gaussian approximation methods. Eder et al. (2025) develop a robust phantom-assisted framework for multi-person localization and vital signs monitoring using MIMO FMCW radar. Terada and Toyoura (2025) introduce a wavelet integrated convolutional neural network for denoising ECG signals acquired with dry electrodes.

Specific applications and optimization problems are addressed in several papers. Tang et al. (2025) investigate dual-function beamforming design for simultaneous multi-target localization and multi-user communication, proposing iterative algorithms based on ADMM and majorization-minimization. Ghavamirad et al. (2025) explore sidelobe level reduction in the ACF of NLFM signals using the smoothing spline method. Chen et al. (2025) analyze NOMA-assisted xURLLC network performance using stochastic network calculus, proposing a power optimization scheme to guarantee stringent KPIs. Munian et al. (2025) present TopoFormer, a hybrid deep learning architecture combining transformers and ConvLSTMs for coastal topography prediction. Mohsin et al. (2025) investigate resource allocation in active RIS-integrated TN-NTN networks using deep reinforcement learning.

Finally, several contributions explore fundamental aspects of signal processing and communication theory. Feng et al. (2025) propose IPP-Net, a deep neural network for indoor pathloss radio map prediction. Mattu et al. (2025) present a low-complexity algorithm for multiple preamble detection in the presence of mobility and delay spread. Dempsey and Ethier (2025) investigate feature representations in CNNs for map-based path loss models. Fejgin and Doclo (2025) propose a method for completing sets of prototype transfer functions for subspace-based DOA estimation of multiple speakers. Bundscherer et al. (2025) develop an ML approach for classifying digital radio operating modes. Haeb-Umbach et al. (2025) compare model-based, data-driven, and hybrid approaches for microphone array signal processing and speech enhancement. Kavianpour et al. (2025) propose a knowledge distillation framework for resource-constrained bearing fault diagnosis. Wang et al. (2025) introduce a diffusion model-based signal detection method that outperforms ML estimation. Hong et al. (2025) study OFDM-FAMA performance in 5G-NR systems. Danish and Grolinger (2025) propose the Kolmogorov-Arnold Recurrent Network for short-term load forecasting. Yildirim et al. (2025) analyze OFDM-based JCAS vulnerabilities in WLAN sensing to spoofing and jamming attacks. Fu et al. (2025) present TWSCardio, a system for cardiac monitoring using in-ear BCG on COTS wireless earbuds. Liu et al. (2025) provide a comprehensive survey of integrated sensing and edge AI (ISEA). Yang et al. (2025) advocate for average reward reinforcement learning in wireless radio resource management. Tao and Sarwate (2025) introduce a differentially private distribution estimation method using functional approximation. Ngo et al. (2025) investigate energy-aware resource allocation for energy harvesting powered wireless sensor nodes. Ahmed et al. (2025) present a survey on advancements in UAV-based ISAC. Zhai et al. (2025) introduce blind multi-mode PMACE for ptychography. Sever et al. (2025) demonstrate a DRL-based dynamic resource allocation xApp using OpenAirInterface. González-Coma et al. (2025) study user selection in near-field gigantic MIMO systems with modular arrays. Güneşer et al. (2025) evaluate RIS optimization algorithms in urban wireless scenarios using Sionna RT. Ouyang et al. (2025) analyze array gain for pinching-antenna systems. Li and Shang (2025) investigate the downlink performance of cell-free massive MIMO for LEO satellite mega-constellations. Zhang et al. (2025) analyze S-band interference from satellite systems on terrestrial networks. Lin et al. (2025) introduce GR-WiFi, a GNU Radio based WiFi platform with MIMO capabilities. Huang et al. (2025) propose a method for distilling calibration via conformalized credal inference. Khodarahmi et al. (2025) present a B1+ mapping technique near metallic implants. Chen et al. (2025) introduce BRIGHT, a globally distributed multimodal building damage assessment dataset. van Nierop et al. (2025) propose deep variational sequential Monte Carlo for high-dimensional observations. Toulemonde et al. (2025) demonstrate enhanced acoustic beamforming with sub-aperture angular multiply and sum. Zheng et al. (2025) investigate UAV swarm-enabled collaborative post-disaster communications. Bereyhi et al. (2025) introduce RegTop-k, a Bayesian framework for gradient sparsification. Gooty et al. (2025) explore precoding design for limited-feedback MISO systems via character-polynomial codes. Vidal Alegría and Edfors (2025) analyze decentralized multi-antenna architectures with unitary constraints. Billat et al. (2025) apply multifractal analysis to physiological signals for optimizing pacing strategy. Li et al. (2025) study the sum rate and user fairness of STAR-RIS aided communications. Schaedler et al. (2025) compare deep neural networks and Volterra series for soft-demapping in short reach optical communication. Li et al. (2025) analyze the coverage and spectral efficiency of NOMA-enabled LEO satellite networks. Sultan et al. (2025) propose optimized sampling for NLOS imaging using modified FFTs.

Paper Highlights

Integrated Sensing and Edge AI: A New Paradigm for Intelligent Perception in 6G

Integrated Sensing and Edge AI: Realizing Intelligent Perception in 6G by Zhiyan Liu, Xu Chen, Hai Wu, Zhanwei Wang, Xianhao Chen, Dusit Niyato, Kaibin Huang https://arxiv.org/abs/2501.06726

Caption: This diagram illustrates the architecture of an Integrated Sensing and Edge AI (ISEA) system, showcasing the interplay between local perception models at edge devices, AirComp pre/post-processing, and global perception models. The system leverages over-the-air computation (AirComp) for efficient data aggregation from distributed sensors, enabling intelligent processing at the network edge. This architecture embodies the principles of modality-aware data communications and integrated communication-computing design discussed in the accompanying text.

Sixth-generation (6G) mobile networks are poised to revolutionize digital interaction, with integrated sensing and edge AI (ISEA) as a key driver. ISEA signifies a paradigm shift, moving beyond simple data transmission to intelligent processing of sensory information at the network edge. This approach tightly couples sensing and AI: sensing provides raw data for AI models, while AI empowers advanced analytics and decision-making from sensory inputs. Edge computing facilitates this symbiotic relationship by bringing AI processing closer to the data source, enabling real-time intelligent operations for applications like autonomous driving, robotics, and smart cities.

ISEA's convergence of sensing, AI, and communication necessitates a holistic design approach that deviates from traditional rate-centric communication paradigms. The paper outlines three core design principles: modality-aware data communications (tailoring transmission protocols to sensory data characteristics), integrated communication-computing design (optimizing protocols for AI-empowered sensing models), and task-oriented optimization (prioritizing sensing task performance). These principles underscore the need for cross-layer synergy and joint optimization across network layers.

ISEA performance evaluation uses combined sensing, communication, and computation metrics. Sensing metrics include accuracy, estimation error, coverage, resolution, and timeliness. Communication metrics encompass throughput, latency, and energy efficiency, while computation metrics focus on time and space complexity. The paper emphasizes the tradeoffs between these metrics, highlighting the "impossible trinity" of sensing performance, communication efficiency, and computation efficiency. Optimizing one often compromises others, requiring careful consideration of application requirements and resource constraints.

The paper surveys ISEA techniques, focusing on how 6G advancements support this paradigm. These include joint source-channel coding (JSCC) (optimizing semantic information transmission), over-the-air computation (AirComp) (leveraging channel superposition for efficient data aggregation), and advanced signal processing (like on-the-fly computing and communication (FlyCom²) and privacy-preserving methods). The paper also explores integrating ISEA with other 6G advancements like XL-MIMO, terahertz communications, RIS, satellite networks, TSN, and DetNet.

Finally, the paper identifies promising future research directions for ISEA, including integrating foundation models, converging ISEA and ISAC, and developing ultra-low-latency ISEA. Foundation models can automate tasks and improve sensing performance. The convergence of ISEA and ISAC offers synergistic enhancements. Ultra-low-latency ISEA is crucial for mission-critical applications, demanding new communication and computation techniques that meet stringent latency requirements while maintaining high sensing accuracy.

BRIGHT: A New Dawn for All-Weather Disaster Response

BRIGHT: A globally distributed multimodal building damage assessment dataset with very-high-resolution for all-weather disaster response by Hongruixuan Chen, Jian Song, Olivier Dietrich, Clifford Broni-Bediako, Weihao Xuan, Junjue Wang, Xinlei Shao, Yimin Wei, Junshi Xia, Cuiling Lan, Konrad Schindler, Naoto Yokoya https://arxiv.org/abs/2501.06019

Disaster events necessitate rapid building damage assessment (BDA) for effective response, where Earth observation (EO) data is crucial. Optical EO data, while widely used, is limited by cloud cover and nighttime conditions. Multimodal EO data, combining optical and SAR imagery, offers a promising all-weather solution. However, developing robust multimodal AI models has been hindered by a lack of suitable datasets. This paper introduces BRIGHT (Building damage assessment dataset using veRy-hIGH-resolution optical and SAR imagery), a groundbreaking open-access, globally distributed, and event-diverse multimodal dataset designed for this challenge.

BRIGHT is the first of its kind, offering a comprehensive resource for AI-based all-weather disaster response. It covers five natural disaster types (earthquakes, storms, wildfires, floods, and volcanic eruptions) and two man-made disasters (explosions and armed conflicts) across 12 diverse regions, focusing on developing countries. It comprises pre-event optical and post-event SAR imagery with very-high resolutions (0.3-1 meter), enabling detailed building-level assessments. Multi-level annotations distinguish between damaged and destroyed buildings, crucial for targeted rescue operations.

Two methodologies were employed and compared: direct semantic segmentation (Y_dam = M_seg(X_T1, X_T2)) and task decoupling (Y_loc = M_loc(X_T1) and Y_clf = M_clf(X_T1, X_T2), with Y_dam = Y_loc ⊙ Y_clf). Seven advanced AI models (UNet, DeepLabV3+, SiamAttnUNet, SiamCRNN, ChangeOS, DamageFormer, and ChangeMamba) were trained and evaluated on BRIGHT.

Results demonstrate the dataset's effectiveness, providing benchmarks for future research. ChangeMamba achieved the best overall performance (OA: 96.65%, mIoU: 67.19%, F_loc: 91.60%, F_clf: 67.93%). DamageFormer also performed well. Earthquake events posed the greatest challenge, highlighting areas for further research. The dataset, code, and trained models are available on GitHub. BRIGHT represents a significant advancement in disaster response, accelerating research in multimodal EO data analysis and contributing to more effective disaster relief.

Erasing Noise: A New Era in Signal Detection with Diffusion Models

Erasing Noise in Signal Detection with Diffusion Model: From Theory to Application by Xiucheng Wang, Peilin Zheng, Nan Cheng https://arxiv.org/abs/2501.07030

Caption: This figure showcases the Symbol Error Rate (SER) performance of a novel diffusion model-based signal detection method compared to Minimum Mean Square Error (MMSE) and Maximum Likelihood (ML) estimation. The proposed method consistently achieves lower SER across various SNR levels, demonstrating its superior performance in noise reduction and signal recovery. The inset graph provides a closer look at the SER performance within a specific SNR range.

This paper introduces a novel signal detection method using denoising diffusion models (DMs), challenging the dominance of Maximum Likelihood (ML) estimation. The authors argue that conventional methods are limited by the assumption of equivalent noise intensity in the received signal and the channel. By reformulating the discrete denoising diffusion probabilistic model (DDPM) into continuous stochastic differential equations (SDEs), they demonstrate the DM's ability to reduce additive Gaussian white noise, potentially surpassing ML estimation.

A key innovation is establishing a mathematical relationship between signal-to-noise ratio (SNR) and the DM's timestep (t). This allows for identifying an optimal t for any given SNR, addressing generalization issues of NN-based methods across varying SNRs. A mathematical scaling technique mitigates out-of-distribution input issues, enabling the trained DM to handle a wide range of SNRs without fine-tuning. The proposed method involves a simple linear operation after noise reduction by the DM. A diffusion transformer (DiT) serves as the backbone neural network (O(n²) complexity). The received signal r is scaled by α:

α = √[(1-t)²Eₛ∼ₛ[||Hs||²] + t] / [Eₛ∼ₛ[||Hs||²] + σ²]

where t is the timestep, H the channel matrix, s the transmitted signal, and σ² the noise density. The DiT predicts noise components hₜ and εₜ, allowing estimation of the transmitted signal Hs:

Hs = αr - thₜ - εₜ

Finally, the transmitted signal s is estimated:

ŝ = H⁻¹Ĥs

Simulations for BPSK and 4QAM show the DM-based method's superiority over ML estimation across various antenna configurations and SNR levels. For example, in a 4x4 MIMO system with BPSK at 5dB SNR, the DM-based method achieves a SER roughly an order of magnitude lower than ML. Similar improvements are observed for 8x8 and 16x16 MIMO systems. The DM-based method also exhibits a steeper SER decline with increasing SNR. Ablation studies confirm the method's robustness and generalization.

The superior performance, especially in high-SNR scenarios, highlights limitations of existing signal detection theory. The proposed framework offers a theoretical advancement, enabling algorithms that surpass classical optimal receiving methods. This is significant for 6G, which demands high data rates and efficient spectrum utilization. Erasing communication noise through intelligent signal processing offers a novel solution to overcome bottlenecks in spectral efficiency and computational complexity. Further research will explore VAE application to handle non-Gaussian noise.

Conclusion

This newsletter showcases a diverse range of cutting-edge research in signal processing and communication. The highlighted papers demonstrate a clear trend towards intelligent, data-driven approaches, particularly within the context of 6G and beyond. The advancements in ISAC, as detailed in the ISEA survey paper, pave the way for a future where wireless networks not only communicate but also perceive and understand their environment. The development of BRIGHT dataset provides a crucial resource for pushing the boundaries of disaster response, enabling more robust and timely assessment of building damage in all-weather conditions. Finally, the groundbreaking work on diffusion models for signal detection demonstrates the potential for surpassing traditional limitations in communication reliability and efficiency, promising significant advancements in spectral efficiency and computational complexity for future wireless systems. These innovations, along with the other contributions covered in the overview, highlight the rapid pace of development in this field and its potential to transform various aspects of our lives.