Adaptive Modulation Research Articles

Accurate and Efficient Event-Based Semantic Segmentation Using Adaptive Spiking Encoder-Decoder Network.

Spiking neural networks (SNNs), known for their low-power, event-driven computation, and intrinsic temporal dynamics, are emerging as promising solutions for processing dynamic, asynchronous signals from event-based sensors. Despite their potential, SNNs face challenges in training and architectural design, resulting in limited performance in challenging event-based dense prediction tasks compared with artificial neural networks (ANNs). In this work, we develop an efficient spiking encoder-decoder network (SpikingEDN) for large-scale event-based semantic segmentation (EbSS) tasks. To enhance the learning efficiency from dynamic event streams, we harness the adaptive threshold which improves network accuracy, sparsity, and robustness in streaming inference. Moreover, we develop a dual-path spiking spatially adaptive modulation (SSAM) module, which is specifically tailored to enhance the representation of sparse events and multimodal inputs, thereby considerably improving network performance. Our SpikingEDN attains a mean intersection over union (MIoU) of 72.57% on the DDD17 dataset and 58.32% on the larger DSEC-Semantic dataset, showing competitive results to the state-of-the-art ANNs while requiring substantially fewer computational resources. Our results shed light on the untapped potential of SNNs in event-based vision applications. The source codes are publicly available at https://github.com/EMI-Group/spikingedn.

MSCS: Multi-stage feature learning with channel-spatial attention mechanism for infrared and visible image fusion

The intention of infrared and visible image fusion is to combine the images captured by different modal sensors in the same scene to enhance its understanding. Deep learning has been proven its powerful application in image fusion due to its fine generalization, robustness, and representability of deep features. However, the performance of these deep learning-based methods heavily depends on the illumination condition. Especially in dark or exposed scenes, the fused results are over-smoothness and low-contrast, resulting in inaccuracy of object detection. To address these issues, this paper develops a multi-stage feature learning approach with channel-spatial attention mechanism, namely MSCS, for infrared and visible image fusion. The MSCS is composed of the following four key procedures: Firstly, the infrared and visible images are decomposed into illumination and reflectance components by a proposed network called as Retinex_Net. Then, the components are transported to an encoder for features coding. Next, we propose an adaptive fusion module with attention mechanisms to fuse the features. Finally, the fused image is generated by the decoder for decoding the fused features. Meanwhile, a novel fusion loss function and a multi-stage training strategy are proposed to train the above modules. The subjective and objective results of experiments on TNO, LLVIP and MSRS datasets illustrate that the proposed method is effective and performs better than the state-of-the-art fusion methods on achieving enjoyable results in dark or over-exposure scenes. And the results of further experiments on the fused images for object detection demonstrate that the fusion outputs produced by our MSCS are more beneficial for detection tasks.

Reducing Noise and Impact of High-Frequency Torque Ripple Caused by Injection Voltages by Using Self-Regulating Random Model Algorithm for SynRMs Sensorless Speed Control

For the sensorless control in a low-speed range of synchronous reluctance motors (SynRMs), injecting random high-frequency (HF) square-wave-type voltages has become a widely used and technologically mature method. It can solve the noise problem of traditional injection signal methods. However, all injection signal methods will cause problems such as torque ripple, which causes speed fluctuations. This article proposes a self-regulating random model algorithm for the random injection signal method, which includes a quantity adaptive module for adding additional random processes, an evaluation module for evaluating torque deviation degree, and an updated model module that is used to receive signals from the other two modules and complete model changes and output random model elements. The main function of this algorithm is to create a model that updates to suppress the evaluation value deviation based on the evaluation situation and outputs an optimal sequence of random numbers, thereby limiting speed bias always in a small range; this can reduce unnecessary changes in the output value of the speed regulator. The feasibility and effectiveness of the proposed algorithm and control method have been demonstrated in experiments based on a 5-kW synchronous reluctance motor.

A portable real-time concrete bridge damage detection system

A drone detection system can achieve real-time inspection of concrete bridge damage. However, it is difficult to deploy existing detection algorithms on a portable system due to their high computational cost. Therefore, a portable damage detection algorithm called CD-YOLOv8 is designed. CD-YOLOv8 is operated on Jetson Xavier NX to detect drone-captured bridge images in real-time. The multi-scale feature adaptive fusion module is used to reduce model parameters and improve detection speed. The spatial context pyramid and deformable convolutional structure are introduced to improve the detection accuracy of tiny concrete bridge damage. Compared with the YOLOv8n, the mAP50 of CD-YOLOv8 is increased by 3.6 %, while the model parameters and FLOPs are decreased by 0.05 M and 0.7 G, respectively. CD-YOLOv8 is also tested on the Jetson Xavier NX with a detection speed of 44 FPS, which is 15.79 % faster than the YOLOv8n. Therefore, CD-YOLOv8 has high accuracy and excellent real-time performance.

A MSA-YOLO Obstacle Detection Algorithm for Rail Transit in Foggy Weather

Obstacles on rail transit significantly compromise operational safety, particularly under dense fog conditions. To address missed and false detections in traditional rail transit detection methods, this paper proposes a multi-scale adaptive YOLO (MSA-YOLO) algorithm. The algorithm incorporates six filters: defog, white balance, gamma, contrast, tone, and sharpen, to remove fog and enhance image quality. However, determining the hyperparameters of these filters is challenging. We employ a multi-scale adaptive module to optimize filter hyperparameters, enhancing fog removal and image quality. Subsequently, YOLO is utilized to detect obstacles on rail transit tracks. The experimental results are encouraging, demonstrating the effectiveness of our proposed method in foggy scenarios.

Accelerating the neural network controller embedded implementation on FPGA with novel dropout techniques for a solar inverter

Accelerating neural network (NN) controllers is important for improving the performance, efficiency, scalability, and reliability of real-time systems, particularly in resource-constrained embedded systems. This paper introduces a novel weight-dropout method for training neural network controllers in real-time closed-loop systems, aimed at accelerating the embedded implementation for solar inverters. The core idea is to eliminate small-magnitude weights during training, thereby reducing the number of necessary connections while ensuring the network’s convergence. To maintain convergence, only non-diagonal elements of the weight matrices were dropped. This dropout technique was integrated into the Levenberg–Marquardt and Forward Accumulation Through Time algorithms, resulting in more efficient training for trajectory tracking. We executed the proposed training algorithm with dropout on the AWS cloud, observing a performance increase of approximately four times compared to local execution. Furthermore, implementing the neural network controller on the Intel Cyclone V Field Programmable Gate Array (FPGA) demonstrates significant improvements in computational and resource efficiency due to the proposed dropout technique leading to sparse weight matrices. This optimization enhances the suitability of the neural network controller for embedded environments. In comparison to Sturtz et al. (2023), which dropped 11 weights, our approach eliminated 18 weights, significantly boosting resource efficiency. This resulted in a 16.40% reduction in Adaptive Logic Modules (ALMs), decreasing the count to 47,426.5. Combinational Look-Up Tables (LUTs) and dedicated logic registers saw reductions of 17.80% and 15.55%, respectively. However, the impact on block memory bits is minimal, showing only a 1% improvement, indicating that memory resources are less affected by weight dropout. In contrast, the usage of Memory 10 Kilobits (MK10s) dropped from 97 to 87, marking a 10% improvement. We also propose an adaptive dropout technique to further improve the previous results.

Hybrid ALM-DSP TDC in Intel Arria 10 FPGA

The article presents the first implementation of a time-to-digital converter (TDC) using a digital signal processing (DSP) block from an Intel field-programmable gate array (FPGA) device. The design and configuration capabilities of the DSP blocks of the leading FPGA manufacturers, AMD and Intel, are compared in the context of time/digital conversion. Next, three variants of TDCs are presented. The problem of ultra-wide bins observed in both the classic version of the converter based on the Adaptive Logic Modules (ALM) and the version built using only DSP blocks has been mitigated in a hybrid variant, combining both solutions. The proposed variant of the hybrid ALM-DSP TDC implemented in Arria 10 FPGA from Intel (20 nm process node) achieved a resolution of 2.1 ps, a single-shot-precision better than 6.36 ps and reduced the maximum differential nonlinearity (DNL) error from 74.5 ps to 12.7 ps.

A multi-scale convolutional neural network with adaptive weight fusion strategy for assisting glaucoma screening

The diagnosis of glaucoma primarily relies on the accurate segmentation of the optic disc (OD) and optic cup (OC). However, OC segmentation remains challenging due to significant individual differences. This study proposes a threshold-based OD extraction method and multi-scale residual U-Net with context attention mechanism and adaptive weight fusion strategy (MRCAAU-Net) to segment OC and OD jointly. Firstly, based on the intensity difference between the OD and the background, an improved thresholding criterion combined with template matching technique is used to effectively detect the OD region, providing accurate input images for network training. Secondly, considering the semantic gap issue during the encoder-decoder feature fusion in the segmentation network, this paper develops an adaptive weight fusion module to guide the effective connections of encoder-decoder features. Additionally, the 1D convolution-based attention mechanism has stronger local perceptiveness. This paper integrate it into various modules, combining multi-scale methods to enhance the segmentation performance of OC by incorporating both global and local features. Finally, the network is guided to learn image segmentation through a hybrid loss function that ignores the background. We conducted extensive experiments on three public datasets, and obtained more accurate segmentation outputs compared to other state-of-the-art methods, particularly in the OC segmentation, where the segmentation accuracy is very close to that of the OD. The proposed joint segmentation network effectively improves and balances the segmentation performance of two objectives, which can greatly assist in large-scale screening for glaucoma.

EAAC-Net: An Efficient Adaptive Attention and Convolution Fusion Network for Skin Lesion Segmentation.

Accurate segmentation of skin lesions in dermoscopic images is of key importance for quantitative analysis of melanoma. Although existing medical image segmentation methods significantly improve skin lesion segmentation, they still have limitations in extracting local features with global information, do not handle challenging lesions well, and usually have a large number of parameters and high computational complexity. To address these issues, this paper proposes an efficient adaptive attention and convolutional fusion network for skin lesion segmentation (EAAC-Net). We designed two parallel encoders, where the efficient adaptive attention feature extraction module (EAAM) adaptively establishes global spatial dependence and global channel dependence by constructing the adjacency matrix of the directed graph and can adaptively filter out the least relevant tokens at the coarse-grained region level, thus reducing the computational complexity of the self-attention mechanism. The efficient multiscale attention-based convolution module (EMA⋅C) utilizes multiscale attention for cross-space learning of local features extracted from the convolutional layer to enhance the representation of richly detailed local features. In addition, we designed a reverse attention feature fusion module (RAFM) to enhance the effective boundary information gradually. To validate the performance of our proposed network, we compared it with other methods on ISIC 2016, ISIC 2018, and PH2 public datasets, and the experimental results show that EAAC-Net has superior segmentation performance under commonly used evaluation metrics.

Online numerical simulation method based on adaptive unscented Kalman filter for shear wall structures

Hybrid test method is an effective approach to evaluate the seismic performance of engineering structures. Particularly, online numerical simulation method (ONSM) based on unscented Kalman filter (UKF) model updating significantly improves the accuracy of the hybrid test results. However, the sensitivity of UKF to the initial values of the state vector, state covariance, and observation noise covariance may lead to the distortion of hybrid test results. To address these issues, one novel ONSM based on adaptive UKF (ONSM-AUKF) is proposed in this paper. The AUKF algorithm incorporates an adaptive variance module to maintain equilibrium. This not only reduces the sensitivity of UKF to the initial parameter settings, prevents filter divergence, but also enhances the accuracy and stability of estimation. The proposed ONSM-AUKF accurately identifies the constitutive model parameters by utilizing measured data from specimens through the AUKF algorithm, which improves the accuracy and the stability of parameter estimation. The simulation results indicate that AUKF algorithm has the capability to decrease the sensitivity of initial parameter settings, thus enhancing the stability and robustness of parameter estimation in comparison to the UKF algorithm. The experimental results validate the feasibility and effectiveness of the proposed ONSM-AUKF, as well as the advantages of the ONSM-AUKF over the ONSM-UKF. The results indicate that the proposed ONSM-AUKF may have broad application prospect in evaluating the seismic performance of various of complex engineering structures.

Seamless rate adaptation for wide SNR range in FSO systems based on rate compatible modulation.

We propose and experimentally demonstrate a large-range and seamless rate-adaptive free-space optical (FSO) scheme based on rate compatible modulation (RCM). It utilizes a mapping method through weighted summation, enabling an even increase of bit energy. This allows the system to seamlessly adjust the throughput. In addition, the handover mechanism based on rate check and acknowledgment signals ensures a wide coverage of signal-to-noise ratio (SNR) fluctuation. It is worth noting that this scheme does not require any changes to the mapping matrix and decoding algorithm, making it easier to implement in real-world systems. Simulation proves the superiority of the proposed approach in terms of coding rate coverage by applying a fixed mapping matrix compared to a traditional adaptive modulation and coding scheme. Experimental demonstration over a 50 m FSO link verifies that the SNR dynamic range of this scheme is >15 dB with a seamless rate adjustment between 6.7 Gb/s and 53.6 Gb/s and the operability of RCM is not limited by the time scale of turbulence, revealing huge potential for the massive connections in future smart cities.

MFPIDet: improved YOLOV7 architecture based on multi-scale feature fusion for prohibited item detection in complex environment

Prohibited item detection is crucial for the safety of public places. Deep learning, one of the mainstream methods in prohibited item detection tasks, has shown superior performance far beyond traditional prohibited item detection methods. However, most neural network architectures in deep learning still lack sufficient local feature representation ability for overlapping and small targets, and ignore the problem of semantic conflicts caused by direct feature fusion. In this paper, we propose MFPIDet, a novel prohibited item detection neural network architecture based on improved YOLOV7 to achieve reliable prohibited item detection in complex environments. Specifically, a multi-scale attention module (MAM) backbone is proposed to filter the redundant information of target regions and further applied to enhance the local feature representation ability of overlapping objects. Here, to reduce the redundant information of target regions, a squeeze-excitation (SE) block is used to filter the background. Then, aiming at enhancing the feature expression ability of overlapping objects, a multi-scale feature extraction module (MFEM) is designed for local feature representation. In addition, to obtain richer context information, We design an adaptive fusion feature pyramid network (AF-FPN) to combine the adaptive context information fusion module (ACIFM) with the feature fusion module (FFM) to improve the neck structure of YOLOV7. The proposed method is validated on the PIDray dataset, and the tested results showed that our method obtained the highest mAP (68.7%), which is improved by 3.5% than YOLOV7 methods. Our approach provides a new design pattern for prohibited item detection in complex environments and shows the development potential of deep learning in related fields.

RAAFNet: Reverse Attention Adaptive Fusion Network for Large-Scale Point Cloud Semantic Segmentation

Point cloud semantic segmentation is essential for comprehending and analyzing scenes. However, performing semantic segmentation on large-scale point clouds presents challenges, including demanding high memory requirements, a lack of structured data, and the absence of topological information. This paper presents a novel method based on the Reverse Attention Adaptive Fusion network (RAAFNet) for segmenting large-scale point clouds. RAAFNet consists of a reverse attention encoder–decoder module, an adaptive fusion module, and a local feature aggregation module. The reverse attention encoder–decoder module is applied to extract point cloud features at different scales. The adaptive fusion module enhances fine-grained representation within multi-resolution feature maps. Furthermore, a local aggregation classifier is introduced, which aggregates the features of neighboring points to the center point in order to leverage contextual information and enhance the classifier’s perceptual capability. Finally, the predicted labels are generated. Notably, our method excels at extracting point cloud features across different dimensions and produces highly accurate segmentation results. Experimental results on the Semantic3D dataset achieved an overall accuracy of 89.9% and a mIoU of 74.4%.

Driver fatigue detection method based on temporal–spatial adaptive networks and adaptive temporal fusion module

The reduction of traffic accidents by determining the driver’s state through fatigue detection is a worthy research issue. Most of the current fatigue driving detection method fails to fully utilize the temporal features of fatigue. To address this problem, this paper proposes a driver fatigue detection method combined with temporal–spatial adaptive networks (TSNet) and adaptive temporal fusion module (ATFM). First, a frame sequence of T frames is obtained by strided sampling of the input video and data enhancement. Subsequently, the temporal–spatial adaptive module (TSAM) is used as the core module and incorporated into Efficientnet-v2 to construct TSNet, which adaptively extracts temporal features according to different videos, adds attention weights to discriminative spatial and channel features, and fully extracts fatiguing temporal–spatial features of videos. Finally, ATFM is utilized to learn the weights between the fatigue classification scores of each frame in the frame sequence and adaptively fuses the fatigue classification scores of individual frames to obtain fatigue prediction results, increasing the extent of the influence of keyframes on the fatigue prediction results. In this paper, the proposed method achieves an accuracy of 89.42% on the NTHU-DDD dataset, which is better than other state-of-the-art methods, and the number of parameters of the proposed method is 24.70M, which is smaller than most of the methods. Through a series of comparative experiments, TSNet and ATFM alone also outperform models and modules with similar functionality.

Coverage hetnet based picocell and femtocell for uplink condition around building environment with single knife edge method

The development of HetNet (Heterogeneous Network) radio base stations has experienced many developments. This is indicated by the existence of microcells, macrocells, picocells, femtocells, and so on. In this research, the research is aimed at the propagation of user equipment communication systems in uplink conditions with HetNet picocells and femtocells. UE propagation is on a straight path between the building environment. The communication frequency used is 10 GHz. The communication between Tx and Rx is modeled as a diffraction mechanism, AWGN channel, and atmospheric attenuation. The Single Knife Edge (SKE) method is used to model the mechanism. The propagation channel is faced with AWGN (Additive White Gaussian Noise). The analysis of this research includes the SNR value, Adaptive Modulation and Coding (AMC) level, and percentage of communication coverage area. The AMC is based on the use of MCS (Modulation and Code Scheme). Some of the MCS modulations used QPSK, 16 QAM, and 64 QAM. As research results show that the percentage of communication coverage area obtained was gNB1 80.59 percent, gNB2 65.67%, and selection combining HetNet 95.52%.

Dual-branch collaborative Siamese network for visual tracking

This paper introduces a dual-branch collaborative Siamese network architecture, CoSiNet, for visual object tracking. The network has a shallow branch for precise target localization and a deep branch for extracting rich semantic information. It integrates two specialized modules - Channel Attention and Spatial Channel Attention Feature Enhancement - for improved feature extraction and background noise reduction. An Adaptive Fusion Module combines response maps from both branches to create an enriched final response map. Experimental results show our algorithm outperforms several contemporary techniques on four different public datasets (OTB100, OTB50, DTB70, TColor128), demonstrating the competitiveness of CoSiNet.

Adaptive spectrum amplitude modulation method for rolling bearing fault frequency determination

Abstract Signal preprocessing and feature extraction are decisive factors in determining the frequency of bearing faults. The presence of noise interference in the status signal of rolling bearings often hampers accurate fault detection. Although there are various methods for preprocessing vibration signals in rolling bearings, they need further improvement in terms of enhancing fault feature expression and localizing fault frequency bands. This limitation significantly hinders the accuracy of fault frequency determination. In order to enhance the representation of fault information on the frequency spectrum, this study proposes a combined approach that incorporates sparse stacked autoencoder (SSAE), wavelet packet decomposition (WPD), and adaptive spectrum amplitude modulation (ASAM). The resulting method is referred to as SSAE-WPD-ASAM. Firstly, the bearing vibration signal is decomposed by wavelet packet according to the scale and frequency band of the signal. On this basis, the signal reconstruction is realized based on the wavelet packet coefficient and energy distribution in different frequency bands. Secondly, for the whole life cycle signal, the reconstructed signal is self-encoded by sparse stacked autoencoder to achieve dimensionality reduction of the reconstructed signal. Then, the spare reconstructed signal is subjected to ASAM. Finally, through envelope demodulation, peak detection of fault frequency and empirical fault frequency comparison, the specific fault types of rolling bearings are determined. The proposed method is verified by theoretical simulation and three groups of practical experiments. The results show that the proposed method has a significant improvement in diagnostic efficiency and accuracy compared with traditional diagnostic methods.

Exploring heterophily in calibration of graph neural networks

With growing applications in safety-critical domains such as autonomous vehicles and healthcare systems, graph neural networks (GNNs) are expected to provide predictions that are not only accurate but also reliable. Confidence calibration is a crucial task closely related to improve the reliability of machine learning models. Current GNN calibration methods primarily address challenges arising from the non-Euclidean topology structure of graph data and a general trend of under-confidence. However, our empirical studies have shown that these methods struggle with local miscalibrations when community homophily differs from the global trend, such as high-homophily nodes in heterophilous graphs. To tackle this issue, we introduce Adaptive Spectral Temperature Scaling (ASTS), a novel spectral-based calibration method tailored for graphs with varying levels of homophily. On top of the node-wise temperature scaling framework, ASTS incorporates two components: (1) dual-channel graph filters that adaptively integrate similar and dissimilar information from neighborhood, (2) an innovative adaptive edge dropout module that alleviates issues induced by low-degree nodes. These designs ensure that ASTS is local structure-aware, global trend-aware and well-adapted to different graph structures. Our extensive empirical studies confirm ASTS’s superior calibration performance over existing methods across graphs of diverse homophily ratios.

Efficient visual anomaly detection model with adaptive wavelet transform

Existing unsupervised surface anomaly detection models have shown promising results. However, research on lightweight and fast anomaly detection models is relatively scarce, thereby impeding their practical application in industrial scenarios where rapid detection is required. Following the approach of reconstruction-based detection methods, we introduce EawT, an efficient anomaly detection model. EawT incorporates a novel pseudo-anomaly synthesis method and a depthwise separable convolutional structure to enhance network lightweighting. Additionally, considering that some studies indicate that Deep Neural Networks (DNNs) may exhibit preferences for certain frequency components during the learning process, potentially affecting the robustness of learned features, we propose an Adaptive Wavelet Transform module designed to modulate frequency components with transfer difficulties in latent space, thereby allowing for dynamic adjustment of feature information to adapt to various requirements. Furthermore, we propose wavelet loss to assist in the reconstruction of original images. Our method achieves comparable detection accuracy to existing reconstruction-based models on the VisA and BTAD datasets, with a frame rate of 228 FPS on a 3090Ti GPU and a lower parameter count, demonstrating the effectiveness of the proposed approach.

Adaptive Modulation and Coding for Underwater Acoustic OTFS Communications Based on Meta-Learning

Adaptive Modulation and Coding for Underwater Acoustic OTFS Communications Based on Meta-Learning