Convolutional neural network(CNN)with the encoder-decoder structure is popular in medical image segmentation due to its excellent local feature extraction ability but it faces limitations in capturing the global featu...Convolutional neural network(CNN)with the encoder-decoder structure is popular in medical image segmentation due to its excellent local feature extraction ability but it faces limitations in capturing the global feature.The transformer can extract the global information well but adapting it to small medical datasets is challenging and its computational complexity can be heavy.In this work,a serial and parallel network is proposed for the accurate 3D medical image segmentation by combining CNN and transformer and promoting feature interactions across various semantic levels.The core components of the proposed method include the cross window self-attention based transformer(CWST)and multi-scale local enhanced(MLE)modules.The CWST module enhances the global context understanding by partitioning 3D images into non-overlapping windows and calculating sparse global attention between windows.The MLE module selectively fuses features by computing the voxel attention between different branch features,and uses convolution to strengthen the dense local information.The experiments on the prostate,atrium,and pancreas MR/CT image datasets consistently demonstrate the advantage of the proposed method over six popular segmentation models in both qualitative evaluation and quantitative indexes such as dice similarity coefficient,Intersection over Union,95%Hausdorff distance and average symmetric surface distance.展开更多
This study introduces the Smart Exponential-Threshold-Linear with Double Deep Q-learning Network(SETL-DDQN)and an extended Gumbel distribution method,designed to optimize the Contention Window(CW)in IEEE 802.11 networ...This study introduces the Smart Exponential-Threshold-Linear with Double Deep Q-learning Network(SETL-DDQN)and an extended Gumbel distribution method,designed to optimize the Contention Window(CW)in IEEE 802.11 networks.Unlike conventional Deep Reinforcement Learning(DRL)-based approaches for CW size adjustment,which often suffer from overestimation bias and limited exploration diversity,leading to suboptimal throughput and collision performance.Our framework integrates the Gumbel distribution and extreme value theory to systematically enhance action selection under varying network conditions.First,SETL adopts a DDQN architecture(SETL-DDQN)to improve Q-value estimation accuracy and enhance training stability.Second,we incorporate a Gumbel distribution-driven exploration mechanism,forming SETL-DDQN(Gumbel),which employs the extreme value theory to promote diverse action selection,replacing the conventional-greedy exploration that undergoes early convergence to suboptimal solutions.Both models are evaluated through extensive simulations in static and time-varying IEEE 802.11 network scenarios.The results demonstrate that our approach consistently achieves higher throughput,lower collision rates,and improved adaptability,even under abrupt fluctuations in traffic load and network conditions.In particular,the Gumbel-based mechanism enhances the balance between exploration and exploitation,facilitating faster adaptation to varying congestion levels.These findings position Gumbel-enhanced DRL as an effective and robust solution for CW optimization in wireless networks,offering notable gains in efficiency and reliability over existing methods.展开更多
With the rapid advancement of Voice over Internet Protocol(VoIP)technology,speech steganography techniques such as Quantization Index Modulation(QIM)and Pitch Modulation Steganography(PMS)have emerged as significant c...With the rapid advancement of Voice over Internet Protocol(VoIP)technology,speech steganography techniques such as Quantization Index Modulation(QIM)and Pitch Modulation Steganography(PMS)have emerged as significant challenges to information security.These techniques embed hidden information into speech streams,making detection increasingly difficult,particularly under conditions of low embedding rates and short speech durations.Existing steganalysis methods often struggle to balance detection accuracy and computational efficiency due to their limited ability to effectively capture both temporal and spatial features of speech signals.To address these challenges,this paper proposes an Efficient Sliding Window Analysis Network(E-SWAN),a novel deep learning model specifically designed for real-time speech steganalysis.E-SWAN integrates two core modules:the LSTM Temporal Feature Miner(LTFM)and the Convolutional Key Feature Miner(CKFM).LTFM captures long-range temporal dependencies using Long Short-Term Memory networks,while CKFM identifies local spatial variations caused by steganographic embedding through convolutional operations.These modules operate within a sliding window framework,enabling efficient extraction of temporal and spatial features.Experimental results on the Chinese CNV and PMS datasets demonstrate the superior performance of E-SWAN.Under conditions of a ten-second sample duration and an embedding rate of 10%,E-SWAN achieves a detection accuracy of 62.09%on the PMS dataset,surpassing existing methods by 4.57%,and an accuracy of 82.28%on the CNV dataset,outperforming state-of-the-art methods by 7.29%.These findings validate the robustness and efficiency of E-SWAN under low embedding rates and short durations,offering a promising solution for real-time VoIP steganalysis.This work provides significant contributions to enhancing information security in digital communications.展开更多
Considering the three-dimensional(3D) U-Net lacks sufficient local feature extraction for image features and lacks attention to the fusion of high-and low-level features, we propose a new model called 3DMAU-Net based ...Considering the three-dimensional(3D) U-Net lacks sufficient local feature extraction for image features and lacks attention to the fusion of high-and low-level features, we propose a new model called 3DMAU-Net based on the 3D U-Net architecture for liver region segmentation. Our model replaces the last two layers of the 3D U-Net with a sliding window-based multilayer perceptron(SMLP), enabling better extraction of local image features. We also design a high-and low-level feature fusion dilated convolution block that focuses on local features and better supplements the surrounding information of the target region. This block is embedded in the entire encoding process, ensuring that the overall network is not simply downsampling. Before each feature extraction, the input features are processed by the dilated convolution block. We validate our experiments on the liver tumor segmentation challenge 2017(Lits2017) dataset, and our model achieves a Dice coefficient of 0.95, which is an improvement of 0.015 compared to the 3D U-Net model. Furthermore, we compare our results with other segmentation methods, and our model consistently outperforms them.展开更多
如果要把一台Microsoft Windows NT服务器加到现有的NetWare网中,许多管理员会面对相同的难题:怎样不需在所有客户机上安装Windows for Workgroups,Windows 95或其它Windows NT客户机软件。又有谁会愿意占用许多内存去改变客户机软件,...如果要把一台Microsoft Windows NT服务器加到现有的NetWare网中,许多管理员会面对相同的难题:怎样不需在所有客户机上安装Windows for Workgroups,Windows 95或其它Windows NT客户机软件。又有谁会愿意占用许多内存去改变客户机软件,安装更多的驱动程序而只是为了增加一台Windows NT服务器。展开更多
基金National Key Research and Development Program of China,Grant/Award Number:2018YFE0206900China Postdoctoral Science Foundation,Grant/Award Number:2023M731204+2 种基金The Open Project of Key Laboratory for Quality Evaluation of Ultrasound Surgical Equipment of National Medical Products Administration,Grant/Award Number:SMDTKL-2023-1-01The Hubei Province Key Research and Development Project,Grant/Award Number:2023BCB007CAAI-Huawei MindSpore Open Fund。
文摘Convolutional neural network(CNN)with the encoder-decoder structure is popular in medical image segmentation due to its excellent local feature extraction ability but it faces limitations in capturing the global feature.The transformer can extract the global information well but adapting it to small medical datasets is challenging and its computational complexity can be heavy.In this work,a serial and parallel network is proposed for the accurate 3D medical image segmentation by combining CNN and transformer and promoting feature interactions across various semantic levels.The core components of the proposed method include the cross window self-attention based transformer(CWST)and multi-scale local enhanced(MLE)modules.The CWST module enhances the global context understanding by partitioning 3D images into non-overlapping windows and calculating sparse global attention between windows.The MLE module selectively fuses features by computing the voxel attention between different branch features,and uses convolution to strengthen the dense local information.The experiments on the prostate,atrium,and pancreas MR/CT image datasets consistently demonstrate the advantage of the proposed method over six popular segmentation models in both qualitative evaluation and quantitative indexes such as dice similarity coefficient,Intersection over Union,95%Hausdorff distance and average symmetric surface distance.
文摘This study introduces the Smart Exponential-Threshold-Linear with Double Deep Q-learning Network(SETL-DDQN)and an extended Gumbel distribution method,designed to optimize the Contention Window(CW)in IEEE 802.11 networks.Unlike conventional Deep Reinforcement Learning(DRL)-based approaches for CW size adjustment,which often suffer from overestimation bias and limited exploration diversity,leading to suboptimal throughput and collision performance.Our framework integrates the Gumbel distribution and extreme value theory to systematically enhance action selection under varying network conditions.First,SETL adopts a DDQN architecture(SETL-DDQN)to improve Q-value estimation accuracy and enhance training stability.Second,we incorporate a Gumbel distribution-driven exploration mechanism,forming SETL-DDQN(Gumbel),which employs the extreme value theory to promote diverse action selection,replacing the conventional-greedy exploration that undergoes early convergence to suboptimal solutions.Both models are evaluated through extensive simulations in static and time-varying IEEE 802.11 network scenarios.The results demonstrate that our approach consistently achieves higher throughput,lower collision rates,and improved adaptability,even under abrupt fluctuations in traffic load and network conditions.In particular,the Gumbel-based mechanism enhances the balance between exploration and exploitation,facilitating faster adaptation to varying congestion levels.These findings position Gumbel-enhanced DRL as an effective and robust solution for CW optimization in wireless networks,offering notable gains in efficiency and reliability over existing methods.
基金supported in part by the Zhejiang Provincial Natural Science Foundation of China under Grant LQ20F020004in part by the National College Student Innovation and Research Training Program under Grant 202313283002.
文摘With the rapid advancement of Voice over Internet Protocol(VoIP)technology,speech steganography techniques such as Quantization Index Modulation(QIM)and Pitch Modulation Steganography(PMS)have emerged as significant challenges to information security.These techniques embed hidden information into speech streams,making detection increasingly difficult,particularly under conditions of low embedding rates and short speech durations.Existing steganalysis methods often struggle to balance detection accuracy and computational efficiency due to their limited ability to effectively capture both temporal and spatial features of speech signals.To address these challenges,this paper proposes an Efficient Sliding Window Analysis Network(E-SWAN),a novel deep learning model specifically designed for real-time speech steganalysis.E-SWAN integrates two core modules:the LSTM Temporal Feature Miner(LTFM)and the Convolutional Key Feature Miner(CKFM).LTFM captures long-range temporal dependencies using Long Short-Term Memory networks,while CKFM identifies local spatial variations caused by steganographic embedding through convolutional operations.These modules operate within a sliding window framework,enabling efficient extraction of temporal and spatial features.Experimental results on the Chinese CNV and PMS datasets demonstrate the superior performance of E-SWAN.Under conditions of a ten-second sample duration and an embedding rate of 10%,E-SWAN achieves a detection accuracy of 62.09%on the PMS dataset,surpassing existing methods by 4.57%,and an accuracy of 82.28%on the CNV dataset,outperforming state-of-the-art methods by 7.29%.These findings validate the robustness and efficiency of E-SWAN under low embedding rates and short durations,offering a promising solution for real-time VoIP steganalysis.This work provides significant contributions to enhancing information security in digital communications.
基金supported by the Shandong Provincial Natural Science Foundation (Nos.ZR2023MF062 and ZR2021MF115)the Introduction and Cultivation Program for Young Innovative Talents of Universities in Shandong (No.2021QCYY003)。
文摘Considering the three-dimensional(3D) U-Net lacks sufficient local feature extraction for image features and lacks attention to the fusion of high-and low-level features, we propose a new model called 3DMAU-Net based on the 3D U-Net architecture for liver region segmentation. Our model replaces the last two layers of the 3D U-Net with a sliding window-based multilayer perceptron(SMLP), enabling better extraction of local image features. We also design a high-and low-level feature fusion dilated convolution block that focuses on local features and better supplements the surrounding information of the target region. This block is embedded in the entire encoding process, ensuring that the overall network is not simply downsampling. Before each feature extraction, the input features are processed by the dilated convolution block. We validate our experiments on the liver tumor segmentation challenge 2017(Lits2017) dataset, and our model achieves a Dice coefficient of 0.95, which is an improvement of 0.015 compared to the 3D U-Net model. Furthermore, we compare our results with other segmentation methods, and our model consistently outperforms them.
文摘如果要把一台Microsoft Windows NT服务器加到现有的NetWare网中,许多管理员会面对相同的难题:怎样不需在所有客户机上安装Windows for Workgroups,Windows 95或其它Windows NT客户机软件。又有谁会愿意占用许多内存去改变客户机软件,安装更多的驱动程序而只是为了增加一台Windows NT服务器。
