Digital Signal Processing最新文献

Uniform allowance model built on the ordered and disordered features of corresponding points

IF 2.9 3区工程技术 Q2 ENGINEERING, ELECTRICAL & ELECTRONIC

Digital Signal Processing

Pub Date : 2025-04-24 DOI: 10.1016/j.dsp.2025.105256

Jingyu Sun , Yadong Gong , Songhua Li , Chuang Zuo , Zichen Zhao , Jibin Zhao , Hongyao Zhang , Ming Cai

Registration is the basis for visual guidance in automated machining processes. This paper focuses on models which are with similar spatial structures. Using bounding boxes to represent outer contours, we extract sparse feature points from point clouds. In this process, matching results are critically affected by erroneous point pairs. Therefore, this paper introduces the Kullback-Leibler (K-L) divergence into the topography evaluation function. A sequential motion-invariant matrix is added to the function to describe the corresponding relationship. To even out the machining allowance, we propose a fine registration method. It considers minimizing variance in the allowance and tangent distance between corresponding points. Meanwhile, the judge criteria of similarity are proposed. They are based on the Hausdorff-Cosine similarity function. This function accounts for angles between neighboring point normals, reducing misidentification and ensuring correct counterparts are included in calculations. Compared with other algorithms, the method improves accuracy, speed of calculations, and ability to resist Gaussian noise. The resulting model ensures uniform allowance distribution. It's a visual prerequisite for further processing.

{"title":"Uniform allowance model built on the ordered and disordered features of corresponding points","authors":"Jingyu Sun , Yadong Gong , Songhua Li , Chuang Zuo , Zichen Zhao , Jibin Zhao , Hongyao Zhang , Ming Cai","doi":"10.1016/j.dsp.2025.105256","DOIUrl":"10.1016/j.dsp.2025.105256","url":null,"abstract":"<div><div>Registration is the basis for visual guidance in automated machining processes. This paper focuses on models which are with similar spatial structures. Using bounding boxes to represent outer contours, we extract sparse feature points from point clouds. In this process, matching results are critically affected by erroneous point pairs. Therefore, this paper introduces the Kullback-Leibler (K-L) divergence into the topography evaluation function. A sequential motion-invariant matrix is added to the function to describe the corresponding relationship. To even out the machining allowance, we propose a fine registration method. It considers minimizing variance in the allowance and tangent distance between corresponding points. Meanwhile, the judge criteria of similarity are proposed. They are based on the Hausdorff-Cosine similarity function. This function accounts for angles between neighboring point normals, reducing misidentification and ensuring correct counterparts are included in calculations. Compared with other algorithms, the method improves accuracy, speed of calculations, and ability to resist Gaussian noise. The resulting model ensures uniform allowance distribution. It's a visual prerequisite for further processing.</div></div>","PeriodicalId":51011,"journal":{"name":"Digital Signal Processing","volume":"164 ","pages":"Article 105256"},"PeriodicalIF":2.9,"publicationDate":"2025-04-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143870848","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"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

DMPNet: A lightweight remote sensing forest wildfire detection network based on multi-scale heterogeneous attention mechanism and dynamic scaling fusion strategy

IF 2.9 3区工程技术 Q2 ENGINEERING, ELECTRICAL & ELECTRONIC

Digital Signal Processing

Pub Date : 2025-04-22 DOI: 10.1016/j.dsp.2025.105252

Yingping Long , Hongwei Ding , Yuanjing Zhu , Zhijun Yang , Bo Li

Unmanned Aerial Vehicle (UAV)-based remote sensing technology has emerged as a critical tool for forest fire detection. However, existing methods face significant challenges in simultaneously achieving high detection accuracy and real-time performance, particularly in scenarios characterized by multi-scale variations, non-stationary behaviors, and complex occlusions. To address these issues, we propose DMPNet, a lightweight neural network specifically designed for forest fire detection. Unlike conventional approaches that sacrifice accuracy for parameter reduction, DMPNet employs an optimized feature extraction architecture (PCSPNet), which not only preserves robust feature representation but also reduces the number of parameters by 16.67% (approximately 500K) compared to YOLOv8n. Furthermore, the dynamic fusion scaling strategy (DSFNet) is integrated into the network's neck to dynamically adjust the size and weight of feature maps, overcoming the limitations of static fusion. Additionally, DMPNet incorporates a multi-scale heterogeneous attention mechanism (MSHA), which effectively addresses occlusion issues through multi-scale contextual reasoning and cross-scale feature interactions. Experimental results demonstrate that DMPNet achieves an inference speed of 1.2 ms, with mAP50 reaching 86.8% and mAP@50-95 at 59%. When compared to YOLOv10n under similar parameter constraints, DMPNet improves mAP50 by 4.2% and surpasses MobileNet, GhostNet, and EfficientNet by over 20% in accuracy. By balancing model compactness, precision, and efficiency, DMPNet provides a robust solution for real-time forest fire monitoring.

{"title":"DMPNet: A lightweight remote sensing forest wildfire detection network based on multi-scale heterogeneous attention mechanism and dynamic scaling fusion strategy","authors":"Yingping Long , Hongwei Ding , Yuanjing Zhu , Zhijun Yang , Bo Li","doi":"10.1016/j.dsp.2025.105252","DOIUrl":"10.1016/j.dsp.2025.105252","url":null,"abstract":"<div><div>Unmanned Aerial Vehicle (UAV)-based remote sensing technology has emerged as a critical tool for forest fire detection. However, existing methods face significant challenges in simultaneously achieving high detection accuracy and real-time performance, particularly in scenarios characterized by multi-scale variations, non-stationary behaviors, and complex occlusions. To address these issues, we propose DMPNet, a lightweight neural network specifically designed for forest fire detection. Unlike conventional approaches that sacrifice accuracy for parameter reduction, DMPNet employs an optimized feature extraction architecture (PCSPNet), which not only preserves robust feature representation but also reduces the number of parameters by 16.67% (approximately 500K) compared to YOLOv8n. Furthermore, the dynamic fusion scaling strategy (DSFNet) is integrated into the network's neck to dynamically adjust the size and weight of feature maps, overcoming the limitations of static fusion. Additionally, DMPNet incorporates a multi-scale heterogeneous attention mechanism (MSHA), which effectively addresses occlusion issues through multi-scale contextual reasoning and cross-scale feature interactions. Experimental results demonstrate that DMPNet achieves an inference speed of 1.2 ms, with mAP50 reaching 86.8% and mAP@50-95 at 59%. When compared to YOLOv10n under similar parameter constraints, DMPNet improves mAP50 by 4.2% and surpasses MobileNet, GhostNet, and EfficientNet by over 20% in accuracy. By balancing model compactness, precision, and efficiency, DMPNet provides a robust solution for real-time forest fire monitoring.</div></div>","PeriodicalId":51011,"journal":{"name":"Digital Signal Processing","volume":"164 ","pages":"Article 105252"},"PeriodicalIF":2.9,"publicationDate":"2025-04-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143870764","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"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

Optic cup and disc segmentation based on multi-level wavelet subband-assisted learning and hybrid convolution-based dual self-attention

IF 2.9 3区工程技术 Q2 ENGINEERING, ELECTRICAL & ELECTRONIC

Digital Signal Processing

Pub Date : 2025-04-22 DOI: 10.1016/j.dsp.2025.105253

Fuying Wang , Suyu Wang , Shangjie Jin

Glaucoma is one of the three leading causes of blindness globally, with over 21 million patients affected in China alone. In clinical practice, accurate segmentation of the optic cup and optic disc is crucial for ophthalmologists in diagnosing glaucoma. Although significant progress has been made, especially with convolutional neural network-based methods, numerous challenges remain in scenarios such as blurred boundaries and blood vessels overlapping the boundaries. To address these issues, a novel optic cup and optic disc segmentation method based on multi-level wavelet subband-assisted learning and hybrid convolution-based dual self-attention are proposed. Firstly, a backbone network enhanced by multi-level wavelet subband auxiliary learning was designed. By introducing wavelet subband information of different directions and frequency bands at various levels of the backbone network, the network can focus more on key features relevant to the segmentation task and achieve more comprehensive feature extraction. Then, a hybrid convolution-based dual self-attention module is designed, which incorporates a ConvMixer module to extract diverse features, thereby enhancing the network's ability to adapt to different scales and shapes of the optic cups and discs. Subsequently, the features are processed by the dual self-attention module in both spatial and channel dimensions and are then reused for deeper feature extraction using different forms of convolution. Finally, feature map multiplication and skip connections are employed to suppress irrelevant features, amplify valuable ones, and refine the overall segmentation results of the optic cup and optic disc. Experimental results on the Drishti-Gs and RIM-ONE-r3 datasets show that the proposed method outperforms most current mainstream algorithms, and it achieves better segmentation results, especially along the edges of the optic cup and disc regions.

{"title":"Optic cup and disc segmentation based on multi-level wavelet subband-assisted learning and hybrid convolution-based dual self-attention","authors":"Fuying Wang , Suyu Wang , Shangjie Jin","doi":"10.1016/j.dsp.2025.105253","DOIUrl":"10.1016/j.dsp.2025.105253","url":null,"abstract":"<div><div>Glaucoma is one of the three leading causes of blindness globally, with over 21 million patients affected in China alone. In clinical practice, accurate segmentation of the optic cup and optic disc is crucial for ophthalmologists in diagnosing glaucoma. Although significant progress has been made, especially with convolutional neural network-based methods, numerous challenges remain in scenarios such as blurred boundaries and blood vessels overlapping the boundaries. To address these issues, a novel optic cup and optic disc segmentation method based on multi-level wavelet subband-assisted learning and hybrid convolution-based dual self-attention are proposed. Firstly, a backbone network enhanced by multi-level wavelet subband auxiliary learning was designed. By introducing wavelet subband information of different directions and frequency bands at various levels of the backbone network, the network can focus more on key features relevant to the segmentation task and achieve more comprehensive feature extraction. Then, a hybrid convolution-based dual self-attention module is designed, which incorporates a ConvMixer module to extract diverse features, thereby enhancing the network's ability to adapt to different scales and shapes of the optic cups and discs. Subsequently, the features are processed by the dual self-attention module in both spatial and channel dimensions and are then reused for deeper feature extraction using different forms of convolution. Finally, feature map multiplication and skip connections are employed to suppress irrelevant features, amplify valuable ones, and refine the overall segmentation results of the optic cup and optic disc. Experimental results on the Drishti-Gs and RIM-ONE-r3 datasets show that the proposed method outperforms most current mainstream algorithms, and it achieves better segmentation results, especially along the edges of the optic cup and disc regions.</div></div>","PeriodicalId":51011,"journal":{"name":"Digital Signal Processing","volume":"164 ","pages":"Article 105253"},"PeriodicalIF":2.9,"publicationDate":"2025-04-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143870831","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"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

On fast RMS estimation for digital data

IF 2.9 3区工程技术 Q2 ENGINEERING, ELECTRICAL & ELECTRONIC

Digital Signal Processing

Pub Date : 2025-04-19 DOI: 10.1016/j.dsp.2025.105254

Sergiusz Sienkowski , Elżbieta Kawecka , Andrzej Perec

This paper presents a new algorithm for fast evaluating the accuracy of root mean square (RMS) estimation for digital data. The algorithm is developed to accelerate the determination of estimation errors. The new algorithm is based on a non-iterative estimator of the RMS parameter and is equivalent to the classical PQN (CPQN) algorithm, which uses an iterative RMS estimator calculated from samples of a harmonic signal occurring in the presence of a constant component (offset), Gaussian noise, and pseudo-quantization noise (PQN) as proposed by Widrow and Kollár. The developed algorithm was called FPQN (fast PQN) and compared with the CPQN algorithm. The classical quantization noise (CQN) algorithm, which is based on an iterative RMS estimator calculated from quantized signal samples, was also used in the comparative studies. The algorithms were compared based on the mean squared errors of the estimators returned by the algorithms. Additionally, the execution times of the algorithms were measured for comparison. The results show that errors of the new algorithm are comparable to those of the CPQN and CQN algorithms. Simultaneously, the new algorithm is several times faster, and its advantage is enhanced as the number of samples in the measurement window increases. For a window containing 200–1000 samples, the developed algorithm is 18–94 times faster than CPQN and 24–120 times faster than CQN.

{"title":"On fast RMS estimation for digital data","authors":"Sergiusz Sienkowski , Elżbieta Kawecka , Andrzej Perec","doi":"10.1016/j.dsp.2025.105254","DOIUrl":"10.1016/j.dsp.2025.105254","url":null,"abstract":"<div><div>This paper presents a new algorithm for fast evaluating the accuracy of root mean square (RMS) estimation for digital data. The algorithm is developed to accelerate the determination of estimation errors. The new algorithm is based on a non-iterative estimator of the RMS parameter and is equivalent to the classical PQN (CPQN) algorithm, which uses an iterative RMS estimator calculated from samples of a harmonic signal occurring in the presence of a constant component (offset), Gaussian noise, and pseudo-quantization noise (PQN) as proposed by Widrow and Kollár. The developed algorithm was called FPQN (fast PQN) and compared with the CPQN algorithm. The classical quantization noise (CQN) algorithm, which is based on an iterative RMS estimator calculated from quantized signal samples, was also used in the comparative studies. The algorithms were compared based on the mean squared errors of the estimators returned by the algorithms. Additionally, the execution times of the algorithms were measured for comparison. The results show that errors of the new algorithm are comparable to those of the CPQN and CQN algorithms. Simultaneously, the new algorithm is several times faster, and its advantage is enhanced as the number of samples in the measurement window increases. For a window containing 200–1000 samples, the developed algorithm is 18–94 times faster than CPQN and 24–120 times faster than CQN.</div></div>","PeriodicalId":51011,"journal":{"name":"Digital Signal Processing","volume":"164 ","pages":"Article 105254"},"PeriodicalIF":2.9,"publicationDate":"2025-04-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143864912","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"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

Brain topographic map: A visual feature for multi-view fusion design in EEG-based biometrics

IF 2.9 3区工程技术 Q2 ENGINEERING, ELECTRICAL & ELECTRONIC

Digital Signal Processing

Pub Date : 2025-04-18 DOI: 10.1016/j.dsp.2025.105251

Dongdong Li, Zhongliang Zeng, Nan Huang, Zhe Wang, Hai Yang

In recent years, lightweight electroencephalogram (EEG) electronic acquisition devices with less electrodes have garnered increasing attention, leading to diverse applications for consumers, such as biometrics systems. Nevertheless, conventional EEG feature extraction relies on matrix data with low spatial resolution from user-friendly acquisition devices, resulting in the lack of brain region information. To this end, we consider the brain topographic map (BTM) as an image-like feature for EEG representation, which promotes us to design a multi-view fusion learning architecture. The proposed method extracts the power spectral density (PSD) feature and interpolates the PSD feature by the Biharmonic Spline Interpolation algorithm to obtain the BTM data, which are then mapped onto the human scalp to generate the BTM image as the BTM feature, explicitly embodying the connectivity and potential information between brain regions. Finally, we establish a multi-view fusion model to fuse the image-view feature and the tensor-view feature. Extensive experimental results show that the BTM feature is competitive and the proposed multi-view fusion model outperforms other models by 4% to 20% improvement in the Correct Recognition Rate. Our research provides a novel visual feature generation aspect with a multi-view fusion design to build robust EEG-based biometrics systems.

{"title":"Brain topographic map: A visual feature for multi-view fusion design in EEG-based biometrics","authors":"Dongdong Li, Zhongliang Zeng, Nan Huang, Zhe Wang, Hai Yang","doi":"10.1016/j.dsp.2025.105251","DOIUrl":"10.1016/j.dsp.2025.105251","url":null,"abstract":"<div><div>In recent years, lightweight electroencephalogram (EEG) electronic acquisition devices with less electrodes have garnered increasing attention, leading to diverse applications for consumers, such as biometrics systems. Nevertheless, conventional EEG feature extraction relies on matrix data with low spatial resolution from user-friendly acquisition devices, resulting in the lack of brain region information. To this end, we consider the brain topographic map (BTM) as an image-like feature for EEG representation, which promotes us to design a multi-view fusion learning architecture. The proposed method extracts the power spectral density (PSD) feature and interpolates the PSD feature by the Biharmonic Spline Interpolation algorithm to obtain the BTM data, which are then mapped onto the human scalp to generate the BTM image as the BTM feature, explicitly embodying the connectivity and potential information between brain regions. Finally, we establish a multi-view fusion model to fuse the image-view feature and the tensor-view feature. Extensive experimental results show that the BTM feature is competitive and the proposed multi-view fusion model outperforms other models by 4% to 20% improvement in the Correct Recognition Rate. Our research provides a novel visual feature generation aspect with a multi-view fusion design to build robust EEG-based biometrics systems.</div></div>","PeriodicalId":51011,"journal":{"name":"Digital Signal Processing","volume":"164 ","pages":"Article 105251"},"PeriodicalIF":2.9,"publicationDate":"2025-04-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143864257","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"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

PI net: An end-to-end semantic decoding model for EEG signals in perception and imagination tasks PI net：感知和想象任务中的脑电信号端到端语义解码模型

IF 2.9 3区工程技术 Q2 ENGINEERING, ELECTRICAL & ELECTRONIC

Digital Signal Processing

Pub Date : 2025-04-18 DOI: 10.1016/j.dsp.2025.105250

Jinze Tong, Wanzhong Chen

Existing methods for decoding the semantics of perception and imagination based on EEG signals primarily focus on isolated tasks or single modalities, neglecting the similarities between perception and imagination EEG signals. This leads to significant limitations in performance and applicability when data types are limited or lack diversity. To address these issues, this paper proposes a novel model, PI Net, for decoding perception and imagination activities. PI Net fully utilizes the similarities between perception and imagination EEG signals, placing them in the same framework for classification, thereby improving the model's decoding performance and generalizability. PI Net first extracts channel and temporal features from EEG signals as shallow features. Then, the Dynamic Convolution in the Spatial-Temporal-Frequency Domain module dynamically adjusts convolution weights based on the spatial-temporal-frequency domain characteristics of the input EEG signals, enabling adaptive processing of temporal-spatial-spectral features. The Assign Feature Weights module, based on the GAU linear attention mechanism, adaptively increases the weights of important features. Finally, PI Net outputs the prediction results for different categories under various tasks through Fully Connected Classification Output using fully connected layers and Softmax layers. Experimental results on publicly available multimodal perception and imagination datasets show that PI Net achieves an accuracy of 92.85% for binary classification and 19.89% for complex 18-class classification tasks, outperforming other models and demonstrating its superior decoding performance and generalizability.

{"title":"PI net: An end-to-end semantic decoding model for EEG signals in perception and imagination tasks","authors":"Jinze Tong, Wanzhong Chen","doi":"10.1016/j.dsp.2025.105250","DOIUrl":"10.1016/j.dsp.2025.105250","url":null,"abstract":"<div><div>Existing methods for decoding the semantics of perception and imagination based on EEG signals primarily focus on isolated tasks or single modalities, neglecting the similarities between perception and imagination EEG signals. This leads to significant limitations in performance and applicability when data types are limited or lack diversity. To address these issues, this paper proposes a novel model, PI Net, for decoding perception and imagination activities. PI Net fully utilizes the similarities between perception and imagination EEG signals, placing them in the same framework for classification, thereby improving the model's decoding performance and generalizability. PI Net first extracts channel and temporal features from EEG signals as shallow features. Then, the Dynamic Convolution in the Spatial-Temporal-Frequency Domain module dynamically adjusts convolution weights based on the spatial-temporal-frequency domain characteristics of the input EEG signals, enabling adaptive processing of temporal-spatial-spectral features. The Assign Feature Weights module, based on the GAU linear attention mechanism, adaptively increases the weights of important features. Finally, PI Net outputs the prediction results for different categories under various tasks through Fully Connected Classification Output using fully connected layers and Softmax layers. Experimental results on publicly available multimodal perception and imagination datasets show that PI Net achieves an accuracy of 92.85% for binary classification and 19.89% for complex 18-class classification tasks, outperforming other models and demonstrating its superior decoding performance and generalizability.</div></div>","PeriodicalId":51011,"journal":{"name":"Digital Signal Processing","volume":"164 ","pages":"Article 105250"},"PeriodicalIF":2.9,"publicationDate":"2025-04-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143860677","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"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

Key parameters for iterative thresholding-type algorithm with nonconvex regularization 非凸正则化迭代阈值型算法的关键参数

IF 2.9 3区工程技术 Q2 ENGINEERING, ELECTRICAL & ELECTRONIC

Digital Signal Processing

Pub Date : 2025-04-17 DOI: 10.1016/j.dsp.2025.105246

Xin Zhou , Zhen Liu , Haisu Zhang , Zhiyuan Zhao , Yongxiang Liu

Iterative thresholding-type algorithm, as one of the typical methods of compressed sensing (CS) theory, is widely used in sparse recovery field, because of its simple computational process. However, the estimation accuracy and convergence speed achieved by this type of algorithm with a nonconvex regularization, e.g., iterative half thresholding (IHalfT) algorithm, are not satisfactory, which limits its practical application. To improve the performance, a modified algorithm is proposed in this paper. Firstly, a novel non-negative expression is introduced in the algorithm to reduce the gap between the relaxation function and the objective function, which can bring tens of dB estimation accuracy improvement, and the convergence of the modified algorithm is verified. Secondly, the fundamental reasons for the remarkable improvement of performance are discussed and analyzed through theoretical derivation. Thirdly, the applicable conditions are elaborated for the modified algorithm. Finally, extensive experimental results demonstrate the effectiveness of the modified iterative thresholding-type algorithm with nonconvex regularization.

{"title":"Key parameters for iterative thresholding-type algorithm with nonconvex regularization","authors":"Xin Zhou , Zhen Liu , Haisu Zhang , Zhiyuan Zhao , Yongxiang Liu","doi":"10.1016/j.dsp.2025.105246","DOIUrl":"10.1016/j.dsp.2025.105246","url":null,"abstract":"<div><div>Iterative thresholding-type algorithm, as one of the typical methods of compressed sensing (CS) theory, is widely used in sparse recovery field, because of its simple computational process. However, the estimation accuracy and convergence speed achieved by this type of algorithm with a nonconvex regularization, e.g., iterative half thresholding (IHalfT) algorithm, are not satisfactory, which limits its practical application. To improve the performance, a modified algorithm is proposed in this paper. Firstly, a novel non-negative expression is introduced in the algorithm to reduce the gap between the relaxation function and the objective function, which can bring tens of dB estimation accuracy improvement, and the convergence of the modified algorithm is verified. Secondly, the fundamental reasons for the remarkable improvement of performance are discussed and analyzed through theoretical derivation. Thirdly, the applicable conditions are elaborated for the modified algorithm. Finally, extensive experimental results demonstrate the effectiveness of the modified iterative thresholding-type algorithm with nonconvex regularization.</div></div>","PeriodicalId":51011,"journal":{"name":"Digital Signal Processing","volume":"164 ","pages":"Article 105246"},"PeriodicalIF":2.9,"publicationDate":"2025-04-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143860701","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"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

ECM Net - A lightweight neural network for target micro-Doppler feature classification in complex scenarios ECM Net - 用于复杂场景中目标微多普勒特征分类的轻量级神经网络

IF 2.9 3区工程技术 Q2 ENGINEERING, ELECTRICAL & ELECTRONIC

Digital Signal Processing

Pub Date : 2025-04-17 DOI: 10.1016/j.dsp.2025.105247

Cunsuo Pang, Zhaonan Liu, Jiachen Sun, Ding Li

With the rapid increase in the number of UAVs, the demand for radar to achieve automatic recognition and classification has become increasingly evident. Micro-Doppler signals can be an important feature of rotary UAVs and are often used in target classification and recognition. However, as the complexity of real-world scenarios changes, the signal-to-noise ratio decreases and the waveform becomes difficult to distinguish. At the same time, it is of great value to study stable, reliable, and highly feasible methods that meet the real-time requirements of recognition results in real-world systems. For this purpose, this paper studies the micro-Doppler characteristics of different rotary UAVs in complex scenarios. Based on deep learning methods, a lightweight ECM Net model is proposed for micro-Doppler signal classification and recognition. This model designs a brand-new ECM downsampling structure, which achieves multi-scale capture of residual features in complex scenarios by traversing the priority of all pixel associations within a certain range. Compared with Xception, ShuffleNetV2 and other models used for micro-Doppler signal classification in recent years, the overall performance of ECM Net achieves the best average testing accuracy, model parameter amount and computational complexity. At the same time, this paper also designs a P2P attention module that can achieve point-to-point specific activation from three dimensions with almost no increase in computational burden, improving the efficiency of ECM Net operations and effectively enhancing the accuracy of the model. Finally, the superiority of the proposed method was verified through actual dataset, with an accuracy of 97.4% and a time consumption of approximately 0.18 seconds.

{"title":"ECM Net - A lightweight neural network for target micro-Doppler feature classification in complex scenarios","authors":"Cunsuo Pang, Zhaonan Liu, Jiachen Sun, Ding Li","doi":"10.1016/j.dsp.2025.105247","DOIUrl":"10.1016/j.dsp.2025.105247","url":null,"abstract":"<div><div>With the rapid increase in the number of UAVs, the demand for radar to achieve automatic recognition and classification has become increasingly evident. Micro-Doppler signals can be an important feature of rotary UAVs and are often used in target classification and recognition. However, as the complexity of real-world scenarios changes, the signal-to-noise ratio decreases and the waveform becomes difficult to distinguish. At the same time, it is of great value to study stable, reliable, and highly feasible methods that meet the real-time requirements of recognition results in real-world systems. For this purpose, this paper studies the micro-Doppler characteristics of different rotary UAVs in complex scenarios. Based on deep learning methods, a lightweight ECM Net model is proposed for micro-Doppler signal classification and recognition. This model designs a brand-new ECM downsampling structure, which achieves multi-scale capture of residual features in complex scenarios by traversing the priority of all pixel associations within a certain range. Compared with Xception, ShuffleNetV2 and other models used for micro-Doppler signal classification in recent years, the overall performance of ECM Net achieves the best average testing accuracy, model parameter amount and computational complexity. At the same time, this paper also designs a P2P attention module that can achieve point-to-point specific activation from three dimensions with almost no increase in computational burden, improving the efficiency of ECM Net operations and effectively enhancing the accuracy of the model. Finally, the superiority of the proposed method was verified through actual dataset, with an accuracy of 97.4% and a time consumption of approximately 0.18 seconds.</div></div>","PeriodicalId":51011,"journal":{"name":"Digital Signal Processing","volume":"164 ","pages":"Article 105247"},"PeriodicalIF":2.9,"publicationDate":"2025-04-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143860700","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"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 hierarchical classification for DoA estimation using coprime array with sensor location errors

IF 2.9 3区工程技术 Q2 ENGINEERING, ELECTRICAL & ELECTRONIC

Digital Signal Processing

Pub Date : 2025-04-15 DOI: 10.1016/j.dsp.2025.105245

Yuxin Zhang, Huijing Dou

Deep learning-based methods have advantages for general Direction of Arrival (DoA) estimation in the array imperfections. However, existing methods using regression or classification are challenged to balance estimation performance and complexity. The good accuracy and adaptability of these methods is achieved by constructing complex deep models as well as large datasets. In this paper, we propose a novel hierarchical classification framework for multi-DoA estimation with coprime array, hoping to improve the accuracy and maintain the adaptation based on an appropriate complexity in the presence of array sensor location errors. Unlike existing data-driven methods, we use a hierarchical classifier that follows the idea of hierarchical modeling of mapping relationships. This makes it easier to learn the classification, thereby reducing the computational burden. The DoA estimation process is divided into multi-level according to the concept of general to specific direction. The complex classification task can then be divided into hierarchical subtasks. We construct a tree structure as priori to provide hierarchical relationships between labels. By learning the semantic relationships between label vectors, our proposed hierarchical model will provide a high-resolution spatial spectra. Our simulation results demonstrate the superiority of the proposed approach over the existing methods in accuracy, adaptation, and complexity.

{"title":"A novel hierarchical classification for DoA estimation using coprime array with sensor location errors","authors":"Yuxin Zhang, Huijing Dou","doi":"10.1016/j.dsp.2025.105245","DOIUrl":"10.1016/j.dsp.2025.105245","url":null,"abstract":"<div><div>Deep learning-based methods have advantages for general Direction of Arrival (DoA) estimation in the array imperfections. However, existing methods using regression or classification are challenged to balance estimation performance and complexity. The good accuracy and adaptability of these methods is achieved by constructing complex deep models as well as large datasets. In this paper, we propose a novel hierarchical classification framework for multi-DoA estimation with coprime array, hoping to improve the accuracy and maintain the adaptation based on an appropriate complexity in the presence of array sensor location errors. Unlike existing data-driven methods, we use a hierarchical classifier that follows the idea of hierarchical modeling of mapping relationships. This makes it easier to learn the classification, thereby reducing the computational burden. The DoA estimation process is divided into multi-level according to the concept of general to specific direction. The complex classification task can then be divided into hierarchical subtasks. We construct a tree structure as priori to provide hierarchical relationships between labels. By learning the semantic relationships between label vectors, our proposed hierarchical model will provide a high-resolution spatial spectra. Our simulation results demonstrate the superiority of the proposed approach over the existing methods in accuracy, adaptation, and complexity.</div></div>","PeriodicalId":51011,"journal":{"name":"Digital Signal Processing","volume":"163 ","pages":"Article 105245"},"PeriodicalIF":2.9,"publicationDate":"2025-04-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143844528","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"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

Automated posterior myocardial infarction detection from vectorcardiogram and derived vectorcardiogram signals using MVFBSE-EWT method

IF 2.9 3区工程技术 Q2 ENGINEERING, ELECTRICAL & ELECTRONIC

Digital Signal Processing

Pub Date : 2025-04-15 DOI: 10.1016/j.dsp.2025.105244

Sibghatullah Inayatullah Khan , Ram Bilas Pachori

It is important to recognize and treat any sign or symptom of posterior myocardial infarction (PMI) promptly. The delay in the diagnosis of PMI may lead to heart failure. Because the standard 12-lead electrocardiogram (ECG) system does not have additional posterior leads, the PMI detection rate using standard 12-lead ECG is low. To improve the diagnostic performance of 12-lead ECG system, additional posterior leads can be added in the existing system. The addition of extra posterior leads may hamper patient comfort and aids in making cardiac monitoring complex. There exist two approaches to address the aforementioned issue. First approach utilizes Frank lead or vectorcardiogram (VCG), wherein, three signals obtained from seven electrodes have been used to record the cardiac activity. In the second approach, the Dowers inverse transform has been used to get derived VCG (dVCG) signal from the standard 12-lead ECG. In the present article, we have employed both the approaches (VCG and dVCG) to detect the PMI using multivariate Fourier-Bessel series expansion based empirical wavelet transform (MVFBSE-EWT). The entropy and complexity features have been extracted from multivariate Fourier-Bessel intrinsic mode functions (MVFBIMFs). The feature space has been reduced using artificial bee colony (ABC) optimization algorithm. Over the reduced feature set, the performance of three hypertuned classifiers, namely, support vector machine (SVM), k-nearest neighbor (KNN), and decision tree (DT) has been compared. The KNN classifier with group k-fold cross-validation strategy proves to be effective in classifying PMI and healthy control (HC) subjects for VCG and dVCG signals with an accuracy of 99.69 % and 99.55 %, respectively. Thus, the proposed method has the potential to enhance PMI detection accuracy without compromising patient comfort, promising practical improvements in clinical diagnostics.

{"title":"Automated posterior myocardial infarction detection from vectorcardiogram and derived vectorcardiogram signals using MVFBSE-EWT method","authors":"Sibghatullah Inayatullah Khan , Ram Bilas Pachori","doi":"10.1016/j.dsp.2025.105244","DOIUrl":"10.1016/j.dsp.2025.105244","url":null,"abstract":"<div><div>It is important to recognize and treat any sign or symptom of posterior myocardial infarction (PMI) promptly. The delay in the diagnosis of PMI may lead to heart failure. Because the standard 12-lead electrocardiogram (ECG) system does not have additional posterior leads, the PMI detection rate using standard 12-lead ECG is low. To improve the diagnostic performance of 12-lead ECG system, additional posterior leads can be added in the existing system. The addition of extra posterior leads may hamper patient comfort and aids in making cardiac monitoring complex. There exist two approaches to address the aforementioned issue. First approach utilizes Frank lead or vectorcardiogram (VCG), wherein, three signals obtained from seven electrodes have been used to record the cardiac activity. In the second approach, the Dowers inverse transform has been used to get derived VCG (dVCG) signal from the standard 12-lead ECG. In the present article, we have employed both the approaches (VCG and dVCG) to detect the PMI using multivariate Fourier-Bessel series expansion based empirical wavelet transform (MVFBSE-EWT). The entropy and complexity features have been extracted from multivariate Fourier-Bessel intrinsic mode functions (MVFBIMFs). The feature space has been reduced using artificial bee colony (ABC) optimization algorithm. Over the reduced feature set, the performance of three hypertuned classifiers, namely, support vector machine (SVM), k-nearest neighbor (KNN), and decision tree (DT) has been compared. The KNN classifier with group k-fold cross-validation strategy proves to be effective in classifying PMI and healthy control (HC) subjects for VCG and dVCG signals with an accuracy of 99.69 % and 99.55 %, respectively. Thus, the proposed method has the potential to enhance PMI detection accuracy without compromising patient comfort, promising practical improvements in clinical diagnostics<em>.</em></div></div>","PeriodicalId":51011,"journal":{"name":"Digital Signal Processing","volume":"163 ","pages":"Article 105244"},"PeriodicalIF":2.9,"publicationDate":"2025-04-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143850452","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"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