Signal Processing最新文献

英文中文

Optimized cone-shaped antenna deployment using rational non-integer array theory 利用合理非整数阵列理论优化锥形天线部署

IF 3.6 2区工程技术 Q2 ENGINEERING, ELECTRICAL & ELECTRONIC

Signal Processing

Pub Date : 2025-12-09 DOI: 10.1016/j.sigpro.2025.110445

Roopesh Kumar Polaganga, Qilian Liang

This study presents a novel exploration of cone-shaped antenna arrays for cellular network deployment, leveraging non-integer array configurations to enhance overall capacity and spectral efficiency while reducing infrastructure costs. Traditional cone-shaped arrays, employing integer-spaced designs, often require a higher density of antenna elements, resulting in increased costs for materials, installation, and maintenance. In this study, we examine two key configurations: the Regular Uniform Cone Array (RUCA) and the Tilted Uniform Cone Array (TUCA), extending both to non-integer designs. By introducing rational spacing along horizontal and vertical dimensions, the non-integer configurations significantly reduce the number of antenna elements required while maintaining or improving critical performance metrics such as sum-rate capacity and spectral efficiency bounds. Through detailed mathematical formulations and MATLAB-based simulations, we demonstrate that non-integer cone-shaped arrays provide improved sum-rate capacity compared to traditional integer-based designs across multiple network configurations. These improvements are particularly significant in next-generation cellular networks, where precise beam control and cost-effective deployment are critical to meeting the high-frequency demands of 6 G. The results underscore the transformative potential of non-integer array configurations in cone-shaped geometries, offering a scalable and efficient solution for future wireless communication infrastructure.

本研究提出了蜂窝网络部署的锥形天线阵列的新探索，利用非整数阵列配置来提高整体容量和频谱效率，同时降低基础设施成本。采用整数间距设计的传统锥形阵列通常需要更高密度的天线元件，从而导致材料、安装和维护成本的增加。在本研究中，我们研究了两种关键配置：规则均匀锥阵列（RUCA）和倾斜均匀锥阵列（TUCA），并将两者扩展到非整数设计。通过在水平和垂直尺寸上引入合理的间距，非整数配置显著减少了所需天线元件的数量，同时保持或提高了关键性能指标，如和速率容量和频谱效率界限。通过详细的数学公式和基于matlab的模拟，我们证明了与传统的基于整数的设计相比，非整数锥形阵列在多种网络配置中提供了更好的和速率容量。这些改进在下一代蜂窝网络中尤为重要，因为精确的波束控制和经济高效的部署对于满足6g的高频需求至关重要。研究结果强调了锥形几何形状的非整数阵列配置的变革潜力，为未来的无线通信基础设施提供了可扩展和高效的解决方案。

{"title":"Optimized cone-shaped antenna deployment using rational non-integer array theory","authors":"Roopesh Kumar Polaganga, Qilian Liang","doi":"10.1016/j.sigpro.2025.110445","DOIUrl":"10.1016/j.sigpro.2025.110445","url":null,"abstract":"<div><div>This study presents a novel exploration of cone-shaped antenna arrays for cellular network deployment, leveraging non-integer array configurations to enhance overall capacity and spectral efficiency while reducing infrastructure costs. Traditional cone-shaped arrays, employing integer-spaced designs, often require a higher density of antenna elements, resulting in increased costs for materials, installation, and maintenance. In this study, we examine two key configurations: the Regular Uniform Cone Array (RUCA) and the Tilted Uniform Cone Array (TUCA), extending both to non-integer designs. By introducing rational spacing along horizontal and vertical dimensions, the non-integer configurations significantly reduce the number of antenna elements required while maintaining or improving critical performance metrics such as sum-rate capacity and spectral efficiency bounds. Through detailed mathematical formulations and MATLAB-based simulations, we demonstrate that non-integer cone-shaped arrays provide improved sum-rate capacity compared to traditional integer-based designs across multiple network configurations. These improvements are particularly significant in next-generation cellular networks, where precise beam control and cost-effective deployment are critical to meeting the high-frequency demands of 6 G. The results underscore the transformative potential of non-integer array configurations in cone-shaped geometries, offering a scalable and efficient solution for future wireless communication infrastructure.</div></div>","PeriodicalId":49523,"journal":{"name":"Signal Processing","volume":"242 ","pages":"Article 110445"},"PeriodicalIF":3.6,"publicationDate":"2025-12-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145790809","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Underwater acoustic signal denoising with diffusion-based generative models 基于扩散生成模型的水声信号去噪

IF 3.6 2区工程技术 Q2 ENGINEERING, ELECTRICAL & ELECTRONIC

Signal Processing

Pub Date : 2025-12-09 DOI: 10.1016/j.sigpro.2025.110430

Boqing Zhu , Yanxin Ma , Zemin Zhou, Wei Guo, Jiahua Zhu, Xiaoqian Zhu

In the work, we proposing a novel generative denoising framework for underwater acoustic denoising based on the diffusion model. Underwater acoustic signal denoising is a challenging task due to the complex, non-stationary, and often non-Gaussian nature of ambient ocean noise. Unlike conventional deep learning approaches that rely heavily on supervised learning and prior knowledge of noise distributions, our method leverages a score-based diffusion model formulated through stochastic differential equations, enabling purely generative training without explicit noise assumptions. Furthermore, we extend the diffusion process and score-matching objective into the complex domain to incorporate phase information, which is essential for reconstructing high-fidelity underwater signals. Extensive experiments on real-world underwater datasets under both simulated and real ambient noise demonstrate the superiority and generalization ability of our approach compared to existing methods.

本文提出了一种新的基于扩散模型的水声去噪生成框架。由于海洋环境噪声的复杂性、非平稳性和非高斯性，水声信号去噪是一项具有挑战性的任务。与传统的深度学习方法严重依赖于监督学习和噪声分布的先验知识不同，我们的方法利用了一个基于分数的扩散模型，该模型是通过随机微分方程制定的，可以在没有明确噪声假设的情况下进行纯粹的生成训练。此外，我们将扩散过程和分数匹配目标扩展到复域，以纳入相位信息，这是重建高保真水下信号所必需的。在模拟和真实环境噪声下对真实水下数据集进行的大量实验表明，与现有方法相比，我们的方法具有优越性和泛化能力。

引用次数: 0

A survey on deep learning enabled automatic modulation classification methods: Data representations, model structures, and regularization techniques 深度学习自动调制分类方法综述：数据表示、模型结构和正则化技术

IF 3.6 2区工程技术 Q2 ENGINEERING, ELECTRICAL & ELECTRONIC

Signal Processing

Pub Date : 2025-12-09 DOI: 10.1016/j.sigpro.2025.110444

Xinyu Tian , Qinghe Zheng , Binglin Li , Dali Qiao , Kan Yu , Zhiqing Wei , Bin Li , Hao Jiang , Xingwang Li , Yun Lin , Guan Gui

Nowadays, intelligent wireless communications have sparked developments in multiple fields due to their ultra-high speed, low latency, and large-scale connectivity capabilities. As a key technique in cognitive communications, automatic modulation classification (AMC) aims to identify the modulation scheme of unknown received signals. AMC has played an important role in both military and civilian applications. Besides, the rapid development of artificial intelligence algorithms represented by deep learning (DL) has brought new opportunities to AMC. In this survey, we investigated a series of DL enabled AMC methods, including key technology, performance, advantages, challenges, and future key development directions. The technical details of various AMC methods are introduced, such as data representation, model structure, and regularization technique in the training process. Extensive experimental results of state-of-the-art DL enabled AMC methods on public or simulated datasets have been compared and analyzed. Despite the achievements that have been made, there are still limitations of existing methods, including generalization capability, inference efficiency, model complexity, and robustness to changing communication parameters. Finally, we have summarized the main challenges faced by DL enabled AMC methods and key future research directions. Critical theoretical foundations and technical routes are envisioned to stimulate core ideas for improving the AMC performance.

如今，智能无线通信因其超高速、低延迟和大规模连接能力而引发了多个领域的发展。自动调制分类（AMC）是认知通信中的一项关键技术，其目的是识别未知接收信号的调制方式。AMC在军事和民用领域都发挥了重要作用。此外，以深度学习（deep learning， DL）为代表的人工智能算法的快速发展也给AMC带来了新的机遇。在本次调查中，我们研究了一系列基于深度学习的AMC方法，包括关键技术、性能、优势、挑战和未来的关键发展方向。介绍了各种AMC方法的技术细节，如数据表示、模型结构、训练过程中的正则化技术等。对公共或模拟数据集上最先进的DL启用AMC方法的大量实验结果进行了比较和分析。尽管已经取得了一些成就，但现有方法仍然存在局限性，包括泛化能力、推理效率、模型复杂性以及对通信参数变化的鲁棒性等。最后，总结了基于深度学习的AMC方法面临的主要挑战和未来的重点研究方向。设想了关键的理论基础和技术路线，以激发提高AMC性能的核心思想。

{"title":"A survey on deep learning enabled automatic modulation classification methods: Data representations, model structures, and regularization techniques","authors":"Xinyu Tian , Qinghe Zheng , Binglin Li , Dali Qiao , Kan Yu , Zhiqing Wei , Bin Li , Hao Jiang , Xingwang Li , Yun Lin , Guan Gui","doi":"10.1016/j.sigpro.2025.110444","DOIUrl":"10.1016/j.sigpro.2025.110444","url":null,"abstract":"<div><div>Nowadays, intelligent wireless communications have sparked developments in multiple fields due to their ultra-high speed, low latency, and large-scale connectivity capabilities. As a key technique in cognitive communications, automatic modulation classification (AMC) aims to identify the modulation scheme of unknown received signals. AMC has played an important role in both military and civilian applications. Besides, the rapid development of artificial intelligence algorithms represented by deep learning (DL) has brought new opportunities to AMC. In this survey, we investigated a series of DL enabled AMC methods, including key technology, performance, advantages, challenges, and future key development directions. The technical details of various AMC methods are introduced, such as data representation, model structure, and regularization technique in the training process. Extensive experimental results of state-of-the-art DL enabled AMC methods on public or simulated datasets have been compared and analyzed. Despite the achievements that have been made, there are still limitations of existing methods, including generalization capability, inference efficiency, model complexity, and robustness to changing communication parameters. Finally, we have summarized the main challenges faced by DL enabled AMC methods and key future research directions. Critical theoretical foundations and technical routes are envisioned to stimulate core ideas for improving the AMC performance.</div></div>","PeriodicalId":49523,"journal":{"name":"Signal Processing","volume":"242 ","pages":"Article 110444"},"PeriodicalIF":3.6,"publicationDate":"2025-12-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145737725","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Graph sampling on a union of periodic graph spectrum subspaces 周期图谱子空间并集上的图采样

IF 3.6 2区工程技术 Q2 ENGINEERING, ELECTRICAL & ELECTRONIC

Signal Processing

Pub Date : 2025-12-07 DOI: 10.1016/j.sigpro.2025.110438

Deyun Wei, Fangdelong Du

Graph signals emerging from a union of subspaces pose significant challenges for existing graph sampling methods. In this paper, firstly, we propose a union of periodic graph spectrum (UPGS) subspaces. The UPGS signal model can represent complex graph signals, such as multiband graph signals with unknown graph spectral support. Secondly, a graph continuous-to-finite (GCTF) block is established based on a new criterion to enable the identification of the non-zero graph spectral support for UPGS signals. Building on this, we present a sampling and recovery framework that incorporates a compressed sampling operator to reduce the required sampling rate. It achieves accurate reconstruction of UPGS signals. In addition, extensive simulations on different graphs demonstrate that our proposed framework achieves superior performance, yielding significantly lower mean square error (MSE) and higher success rates compared to alternative methods. Finally, our method can perform graph signal detection with high accuracy.

从子空间并集中产生的图信号对现有的图采样方法提出了重大挑战。本文首先提出了周期图谱子空间的并集。UPGS信号模型可以表示复杂的图信号，如具有未知图谱支持的多频带图信号。其次，建立了基于新准则的图连续到有限（GCTF）块，实现了UPGS信号非零图谱支持度的识别；在此基础上，我们提出了一个采样和恢复框架，该框架包含压缩采样算子以降低所需的采样率。实现了UPGS信号的精确重构。此外，在不同图形上的大量模拟表明，与其他方法相比，我们提出的框架实现了卓越的性能，产生了显着更低的均方误差（MSE）和更高的成功率。最后，我们的方法可以实现高精度的图形信号检测。

引用次数: 0

New sufficient condition of ℓ1−βℓq method for robust signal recovery 新的1 - β q鲁棒信号恢复方法的充分条件

IF 3.6 2区工程技术 Q2 ENGINEERING, ELECTRICAL & ELECTRONIC

Signal Processing

Pub Date : 2025-12-07 DOI: 10.1016/j.sigpro.2025.110435

Jing Zhang , Wendong Wang

The problem of sparse signal recovery has drawn widespread attention from researchers. Based on high-order restricted isometry property (RIP) analysis, this paper establishes a new sufficient condition for

ℓ_{1} - β ℓ_{q}

method to guarantee robust signal recovery. This condition is shown to be better than the existing ones. Notably, when specialized to

ℓ_{1} - ℓ_{2}

method, our derived condition yields a better upper bound on RIC compared to the state-of-the-art ones. Furthermore, our analysis provides a theoretical perspective that extends the recovery guarantees of the convex ℓ₁ method to the non-convex

ℓ_{1} - β ℓ_{q}

method.

稀疏信号恢复问题引起了研究者的广泛关注。基于高阶受限等距特性（RIP）分析，建立了保证信号鲁棒恢复的一个新的充分条件。该条件优于现有条件。值得注意的是，当专门化到1−2方法时，我们推导的条件在RIC上得到了比现有条件更好的上界。此外，我们的分析提供了一个理论视角，将凸1方法的恢复保证扩展到非凸1 - β 1 q方法。

引用次数: 0

Reversible data hiding for color images based on a frequency-first partial assignment strategy 基于频率优先部分分配策略的彩色图像可逆数据隐藏

IF 3.6 2区工程技术 Q2 ENGINEERING, ELECTRICAL & ELECTRONIC

Signal Processing

Pub Date : 2025-12-04 DOI: 10.1016/j.sigpro.2025.110429

Wei-Chun Lin , Tsai-Ju Lee , Hsin-Lung Wu

This paper investigates adaptive reversible data hiding (RDH) schemes for color images based on optimal 3D-mapping selection. A typical 3D-mapping selection scheme consists of two main steps. First, a histogram of 3D-prediction-error vectors is generated for a given color image using a pixel predictor. Then, a reversible 3D-mapping is determined through a selection mechanism. This study focuses primarily on the construction of the 3D-mapping selection mechanism. Previous approaches include iteratively modifying a basic 3D mapping, employing deep reinforcement learning to search for an optimal 3D mapping, and constructing a 3D mapping by optimizing 2D mappings. However, these state-of-the-art selection schemes often suffer from either high computational cost or limited search space. To enhance embedding performance, this paper proposes a novel 3D-mapping selection framework based on a frequency-first heuristic partial assignment strategy. In this strategy, 3D prediction error vectors with higher histogram frequencies are allocated a larger predefined mapping range. This method effectively reduces the search space and results in an efficient RDH algorithm. Comprehensive experimental analysis demonstrates that the proposed method outperforms existing adaptive histogram modification-based RDH approaches on most test color images.

研究了基于最优三维映射选择的彩色图像自适应可逆数据隐藏（RDH）方案。典型的3d贴图选择方案包括两个主要步骤。首先，使用像素预测器为给定的彩色图像生成3d预测误差向量的直方图。然后，通过选择机制确定可逆的3d映射。本研究主要围绕三维映射选择机制的构建展开。以前的方法包括迭代修改基本的3D映射，使用深度强化学习来搜索最优的3D映射，以及通过优化2D映射来构建3D映射。然而，这些最先进的选择方案往往存在计算成本高或搜索空间有限的问题。为了提高嵌入性能，本文提出了一种基于频率优先启发式部分分配策略的三维映射选择框架。在该策略中，直方图频率较高的3D预测误差向量被分配到更大的预定义映射范围。该方法有效地减小了搜索空间，得到了一种高效的RDH算法。综合实验分析表明，该方法在大多数测试彩色图像上优于现有的基于自适应直方图修改的RDH方法。

{"title":"Reversible data hiding for color images based on a frequency-first partial assignment strategy","authors":"Wei-Chun Lin , Tsai-Ju Lee , Hsin-Lung Wu","doi":"10.1016/j.sigpro.2025.110429","DOIUrl":"10.1016/j.sigpro.2025.110429","url":null,"abstract":"<div><div>This paper investigates adaptive reversible data hiding (RDH) schemes for color images based on optimal 3D-mapping selection. A typical 3D-mapping selection scheme consists of two main steps. First, a histogram of 3D-prediction-error vectors is generated for a given color image using a pixel predictor. Then, a reversible 3D-mapping is determined through a selection mechanism. This study focuses primarily on the construction of the 3D-mapping selection mechanism. Previous approaches include iteratively modifying a basic 3D mapping, employing deep reinforcement learning to search for an optimal 3D mapping, and constructing a 3D mapping by optimizing 2D mappings. However, these state-of-the-art selection schemes often suffer from either high computational cost or limited search space. To enhance embedding performance, this paper proposes a novel 3D-mapping selection framework based on a frequency-first heuristic partial assignment strategy. In this strategy, 3D prediction error vectors with higher histogram frequencies are allocated a larger predefined mapping range. This method effectively reduces the search space and results in an efficient RDH algorithm. Comprehensive experimental analysis demonstrates that the proposed method outperforms existing adaptive histogram modification-based RDH approaches on most test color images.</div></div>","PeriodicalId":49523,"journal":{"name":"Signal Processing","volume":"242 ","pages":"Article 110429"},"PeriodicalIF":3.6,"publicationDate":"2025-12-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145737789","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

A novel distance-velocity estimation algorithm for FMCW-LiDAR based on trapezoid wave 基于梯形波的FMCW-LiDAR距离-速度估计新算法

IF 3.6 2区工程技术 Q2 ENGINEERING, ELECTRICAL & ELECTRONIC

Signal Processing

Pub Date : 2025-12-02 DOI: 10.1016/j.sigpro.2025.110434

Wenyan Hong , Shiyi Shen , Shihuang Wu , Huiying Li , Mengting Lian , Yixiong Zhang , Caipin Li , Jianyang Zhou

Frequency Modulation Continuous Wave - Light Detection and Ranging (FMCW-LiDAR) can simultaneously estimate the distance and Doppler information of a target in a single measurement. The traditional FMCW-LiDAR employs a triangular wave and I-channel sampling, resulting in a symmetric spectrum. However, when the absolute value of the target Doppler frequency exceeds the distance frequency, there is an issue of distance-velocity coupling. In addition, when the laser beam passes through the edge of a target, the LiDAR may receive two echoes from different targets, making target frequency pairing difficult. In this paper, we propose a novel distance-velocity estimation algorithm based on trapezoid wave. The proposed method utilizes a constant frequency part of trapezoid wave to estimate Doppler frequency. Then, the absolute value of the Doppler frequency is used as a criterion for negative frequency estimation and target matching. Simulation results show that the distance-velocity frequency estimation range of the proposed method is twice as high as that of the triangular wave method and the utilization rate of time resources has increased by 25 % compared with the variable-frequency triangular wave.

调频连续波光探测和测距（FMCW-LiDAR）可以在一次测量中同时估计目标的距离和多普勒信息。传统的FMCW-LiDAR采用三角波和i通道采样，从而产生对称频谱。然而，当目标多普勒频率的绝对值超过距离频率时，就会出现距离-速度耦合问题。此外，当激光束穿过目标边缘时，激光雷达可能接收到来自不同目标的两个回波，使目标频率配对变得困难。本文提出了一种新的基于梯形波的距离-速度估计算法。该方法利用梯形波的恒频部分来估计多普勒频率。然后，将多普勒频率的绝对值作为负频率估计和目标匹配的准则。仿真结果表明，该方法的距离-速度频率估计范围是三角波法的2倍，时间资源利用率比变频三角波法提高了25%。

{"title":"A novel distance-velocity estimation algorithm for FMCW-LiDAR based on trapezoid wave","authors":"Wenyan Hong , Shiyi Shen , Shihuang Wu , Huiying Li , Mengting Lian , Yixiong Zhang , Caipin Li , Jianyang Zhou","doi":"10.1016/j.sigpro.2025.110434","DOIUrl":"10.1016/j.sigpro.2025.110434","url":null,"abstract":"<div><div>Frequency Modulation Continuous Wave - Light Detection and Ranging (FMCW-LiDAR) can simultaneously estimate the distance and Doppler information of a target in a single measurement. The traditional FMCW-LiDAR employs a triangular wave and I-channel sampling, resulting in a symmetric spectrum. However, when the absolute value of the target Doppler frequency exceeds the distance frequency, there is an issue of distance-velocity coupling. In addition, when the laser beam passes through the edge of a target, the LiDAR may receive two echoes from different targets, making target frequency pairing difficult. In this paper, we propose a novel distance-velocity estimation algorithm based on trapezoid wave. The proposed method utilizes a constant frequency part of trapezoid wave to estimate Doppler frequency. Then, the absolute value of the Doppler frequency is used as a criterion for negative frequency estimation and target matching. Simulation results show that the distance-velocity frequency estimation range of the proposed method is twice as high as that of the triangular wave method and the utilization rate of time resources has increased by 25 % compared with the variable-frequency triangular wave.</div></div>","PeriodicalId":49523,"journal":{"name":"Signal Processing","volume":"242 ","pages":"Article 110434"},"PeriodicalIF":3.6,"publicationDate":"2025-12-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145737790","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Plug-and-play adaptive rank estimation for low-rank tensor completion 低秩张量补全的即插即用自适应秩估计

IF 3.6 2区工程技术 Q2 ENGINEERING, ELECTRICAL & ELECTRONIC

Signal Processing

Pub Date : 2025-11-29 DOI: 10.1016/j.sigpro.2025.110421

Chenghu Mi, Jingfei He, Ao Li, Xiaotong Liu

Tensor completion methods based on matrix factorization are often used to recover missing data in images. However, when the missing ratio is high, the accuracy of the predefined rank significantly affects the performance of the completion model. To address this issue, this paper proposes a plug-and-play adaptive rank estimation (ARE) framework that automatically estimates the tensor rank for matrix factorization models, applicable to various rank types. Specifically, the framework interpolates the observed image using Delaunay triangulation and generates a structurally complete reference image by filling in missing pixels. A rank estimation strategy based on backtracking search is then introduced, which utilizes the reference image to find an approximately optimal value for each rank component. To better capture potential structural information from observational data, a pseudo-observation fill (POF) module is designed to enrich the available data. Experimental results on four advanced tensor ranks demonstrate that the proposed method can effectively estimate them and enhances recovery performance.

基于矩阵分解的张量补全方法常用于图像缺失数据的恢复。然而，当缺失率较高时，预定义秩的准确性会显著影响补全模型的性能。为了解决这个问题，本文提出了一个即插即用的自适应秩估计（ARE）框架，该框架可自动估计矩阵分解模型的张量秩，适用于各种秩类型。具体而言，该框架使用Delaunay三角剖分法对观测图像进行插值，并通过填充缺失像素生成结构完整的参考图像。然后介绍了一种基于回溯搜索的秩估计策略，该策略利用参考图像找到每个秩分量的近似最优值。为了更好地从观测数据中获取潜在的结构信息，设计了伪观测填充（POF）模块来丰富可用数据。对4个高级张量秩的实验结果表明，该方法能有效地估计张量秩，提高了复原性能。

引用次数: 0

Improved ADTFD-class algorithms for HFM signals based on direction extension using an energy concentration criterion 基于能量集中准则的HFM信号方向扩展改进adtfd类算法

IF 3.6 2区工程技术 Q2 ENGINEERING, ELECTRICAL & ELECTRONIC

Signal Processing

Pub Date : 2025-11-29 DOI: 10.1016/j.sigpro.2025.110422

Shuai Yao , Jinyu Lin , Yixuan Zhang , Xincheng Zhao , Zixu Wang , Qisong Wu , Chaochao Wang

Fast adaptive directional time-frequency distribution (F-ADTFD) represents an efficient variant of ADTFD, achieving a balance between low computational complexity and high-performance time-frequency analysis. However, its direction estimation methodology may lead to the loss of auto-term directions when employed for analyzing hyperbolic frequency modulated (HFM) signals, where auto-term directions exhibit time-varying characteristics. To address this dilemma, a novel direction extension framework guided by an energy concentration criterion for HFM signals is proposed in this paper. The framework operates in two sequential stages: first, identifying potential auto-term directions of the HFM signal in (ν, τ) plane, and second, extending these directions by using the ratio of norms concentration measure in (t, f) plane. Leveraging the aforementioned framework, two specialized time-frequency analysis algorithms are developed by integrating with fast ADTFD (F-ADTFD) and locally optimized ADTFD (LO-ADTFD), respectively, namely improved F-ADTFD (IF-ADTFD) and fast LO-ADTFD (FLO-ADTFD). Both simulation and experimental results have verified that compared with F-ADTFD, IF-ADTFD mitigates the risk of signal component loss during HFM signal processing. Additionally, FLO-ADTFD achieves performance comparable to LO-ADTFD with 9–54 % reduced computational complexity, demonstrating an average reduction of 40 % across simulations and experiment.

快速自适应定向时频分布（F-ADTFD）是ADTFD的一种有效变体，在低计算复杂度和高性能时频分析之间取得了平衡。然而，其方向估计方法在分析双曲调频（HFM）信号时可能导致自动项方向的丢失，其中自动项方向具有时变特性。为了解决这一难题，本文提出了一种基于能量集中准则的高频调频信号方向扩展框架。该框架分为两个连续的阶段：首先，确定（ν, τ）平面中高频调频信号的潜在自动项方向，其次，通过使用（t, f）平面中的规范浓度比测量来扩展这些方向。利用上述框架，通过集成快速ADTFD （F-ADTFD）和局部优化ADTFD (LO-ADTFD)，分别开发了两种专门的时频分析算法，即改进的F-ADTFD （IF-ADTFD）和快速的LO-ADTFD （fl -ADTFD）。仿真和实验结果均证实，与F-ADTFD相比，IF-ADTFD降低了高频调频信号处理过程中信号分量丢失的风险。此外，FLO-ADTFD的性能与LO-ADTFD相当，计算复杂度降低了9 - 54%，在模拟和实验中平均降低了40%。

{"title":"Improved ADTFD-class algorithms for HFM signals based on direction extension using an energy concentration criterion","authors":"Shuai Yao , Jinyu Lin , Yixuan Zhang , Xincheng Zhao , Zixu Wang , Qisong Wu , Chaochao Wang","doi":"10.1016/j.sigpro.2025.110422","DOIUrl":"10.1016/j.sigpro.2025.110422","url":null,"abstract":"<div><div>Fast adaptive directional time-frequency distribution (F-ADTFD) represents an efficient variant of ADTFD, achieving a balance between low computational complexity and high-performance time-frequency analysis. However, its direction estimation methodology may lead to the loss of auto-term directions when employed for analyzing hyperbolic frequency modulated (HFM) signals, where auto-term directions exhibit time-varying characteristics. To address this dilemma, a novel direction extension framework guided by an energy concentration criterion for HFM signals is proposed in this paper. The framework operates in two sequential stages: first, identifying potential auto-term directions of the HFM signal in (<em>ν, τ</em>) plane, and second, extending these directions by using the ratio of norms concentration measure in (<em>t, f</em>) plane. Leveraging the aforementioned framework, two specialized time-frequency analysis algorithms are developed by integrating with fast ADTFD (F-ADTFD) and locally optimized ADTFD (LO-ADTFD), respectively, namely improved F-ADTFD (IF-ADTFD) and fast LO-ADTFD (FLO-ADTFD). Both simulation and experimental results have verified that compared with F-ADTFD, IF-ADTFD mitigates the risk of signal component loss during HFM signal processing. Additionally, FLO-ADTFD achieves performance comparable to LO-ADTFD with 9–54 % reduced computational complexity, demonstrating an average reduction of 40 % across simulations and experiment.</div></div>","PeriodicalId":49523,"journal":{"name":"Signal Processing","volume":"242 ","pages":"Article 110422"},"PeriodicalIF":3.6,"publicationDate":"2025-11-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145684880","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Stride conversion algorithms for convolutional layers and its application to sampling-frequency-independent deep neural networks 卷积层跨步转换算法及其在采样频率无关深度神经网络中的应用

IF 3.6 2区工程技术 Q2 ENGINEERING, ELECTRICAL & ELECTRONIC

Signal Processing

Pub Date : 2025-11-28 DOI: 10.1016/j.sigpro.2025.110420

Kanami Imamura , Tomohiko Nakamura , Norihiro Takamune , Kohei Yatabe , Hiroshi Saruwatari

We propose interpolation-based algorithms that enable convolutional and transposed convolutional layers to operate with arbitrary (including non-integer) strides. A primary motivation for the proposed algorithms is to maintain a consistent temporal resolution when adapting deep neural networks (DNNs) to different sampling frequencies (SFs). To handle untrained SFs, we previously introduced SF-independent (SFI) convolutional layers, which adjust kernel weights in accordance with the target SF. However, achieving full consistency across SFs also requires the proportional adjustment of the stride, which results in non-integer values in many practical cases. Conventional algorithms for convolutional layers cannot handle such strides directly, and commonly used approaches (e.g., stride rounding or signal resampling) lead to performance degradation. To solve this problem, we propose a feature-domain interpolation framework that constructs continuous-time representations of intermediate features. This enables sampling at arbitrary stride intervals without modifying the network architecture. Through music source separation experiments, we show that the proposed algorithms maintain a strong performance across a range of SFs, including those where the stride becomes non-integer. Our analysis reveals that the proposed algorithms are robust to the choice of interpolation method and are especially effective for sources containing pitched sounds.

我们提出了基于插值的算法，使卷积和转置卷积层可以任意（包括非整数）步进操作。所提出算法的主要动机是在使深度神经网络（dnn）适应不同采样频率（sf）时保持一致的时间分辨率。为了处理未经训练的SF，我们之前引入了独立于SF （SFI）的卷积层，它根据目标SF调整核权重。然而，要实现跨SFs的完全一致性还需要按比例调整步幅，这在许多实际情况下会导致非整数值。卷积层的传统算法不能直接处理这种跨步，而常用的方法（例如，跨步舍入或信号重采样）会导致性能下降。为了解决这个问题，我们提出了一个特征域插值框架，该框架构建了中间特征的连续时间表示。这样就可以在不修改网络架构的情况下以任意步幅间隔进行采样。通过音乐源分离实验，我们证明了所提出的算法在一系列sf范围内保持了很强的性能，包括那些跨距变为非整数的sf。我们的分析表明，所提出的算法对插值方法的选择具有鲁棒性，并且对包含音调声音的声源特别有效。

{"title":"Stride conversion algorithms for convolutional layers and its application to sampling-frequency-independent deep neural networks","authors":"Kanami Imamura , Tomohiko Nakamura , Norihiro Takamune , Kohei Yatabe , Hiroshi Saruwatari","doi":"10.1016/j.sigpro.2025.110420","DOIUrl":"10.1016/j.sigpro.2025.110420","url":null,"abstract":"<div><div>We propose interpolation-based algorithms that enable convolutional and transposed convolutional layers to operate with arbitrary (including non-integer) strides. A primary motivation for the proposed algorithms is to maintain a consistent temporal resolution when adapting deep neural networks (DNNs) to different sampling frequencies (SFs). To handle untrained SFs, we previously introduced SF-independent (SFI) convolutional layers, which adjust kernel weights in accordance with the target SF. However, achieving full consistency across SFs also requires the proportional adjustment of the stride, which results in non-integer values in many practical cases. Conventional algorithms for convolutional layers cannot handle such strides directly, and commonly used approaches (e.g., stride rounding or signal resampling) lead to performance degradation. To solve this problem, we propose a feature-domain interpolation framework that constructs continuous-time representations of intermediate features. This enables sampling at arbitrary stride intervals without modifying the network architecture. Through music source separation experiments, we show that the proposed algorithms maintain a strong performance across a range of SFs, including those where the stride becomes non-integer. Our analysis reveals that the proposed algorithms are robust to the choice of interpolation method and are especially effective for sources containing pitched sounds.</div></div>","PeriodicalId":49523,"journal":{"name":"Signal Processing","volume":"242 ","pages":"Article 110420"},"PeriodicalIF":3.6,"publicationDate":"2025-11-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145684879","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

首页上一页

下一页尾页

类型

全部化学•材料生命科学医学物理工程技术环境•农林材料科学地球科学法学管理学化学环境科学与生态学计算机科学教育学经济学农林科学人文科学生物学数学物理与天体物理心理学综合性期刊其他工业工程理学历史学农学文学信息工程

数据库

全部 ACS Publications Elsevier ieeexplore Springer The Royal Society of Chemistry Wiley

期刊

Signal Processing

全部 Acc. Chem. Res. ACS Applied Bio Materials ACS Appl. Electron. Mater. ACS Appl. Energy Mater. ACS Appl. Mater. Interfaces ACS Appl. Nano Mater. ACS Appl. Polym. Mater. ACS BIOMATER-SCI ENG ACS Catal. ACS Cent. Sci. ACS Chem. Biol. ACS Chemical Health & Safety ACS Chem. Neurosci. ACS Comb. Sci. ACS Earth Space Chem. ACS Energy Lett. ACS Infect. Dis. ACS Macro Lett. ACS Mater. Lett. ACS Med. Chem. Lett. ACS Nano ACS Omega ACS Photonics ACS Sens. ACS Sustainable Chem. Eng. ACS Synth. Biol. Anal. Chem. BIOCHEMISTRY-US Bioconjugate Chem. BIOMACROMOLECULES Chem. Res. Toxicol. Chem. Rev. Chem. Mater. CRYST GROWTH DES ENERG FUEL Environ. Sci. Technol. Environ. Sci. Technol. Lett. Eur. J. Inorg. Chem. IND ENG CHEM RES Inorg. Chem. J. Agric. Food. Chem. J. Chem. Eng. Data J. Chem. Educ. J. Chem. Inf. Model. J. Chem. Theory Comput. J. Med. Chem. J. Nat. Prod. J PROTEOME RES J. Am. Chem. Soc. LANGMUIR MACROMOLECULES Mol. Pharmaceutics Nano Lett. Org. Lett. ORG PROCESS RES DEV ORGANOMETALLICS J. Org. Chem. J. Phys. Chem. J. Phys. Chem. A J. Phys. Chem. B J. Phys. Chem. C J. Phys. Chem. Lett. Analyst Anal. Methods Biomater. Sci. Catal. Sci. Technol. Chem. Commun. Chem. Soc. Rev. CHEM EDUC RES PRACT CRYSTENGCOMM Dalton Trans. Energy Environ. Sci. ENVIRON SCI-NANO ENVIRON SCI-PROC IMP ENVIRON SCI-WAT RES Faraday Discuss. Food Funct. Green Chem. Inorg. Chem. Front. Integr. Biol. J. Anal. At. Spectrom. J. Mater. Chem. A J. Mater. Chem. B J. Mater. Chem. C Lab Chip Mater. Chem. Front. Mater. Horiz. MEDCHEMCOMM Metallomics Mol. Biosyst. Mol. Syst. Des. Eng. Nanoscale Nanoscale Horiz. Nat. Prod. Rep. New J. Chem. Org. Biomol. Chem. Org. Chem. Front. PHOTOCH PHOTOBIO SCI PCCP Polym. Chem.

﹀