Multi-channel Feature Research Articles

A multilevel cooperative attention network of precise quantitative analysis for predicting ractopamine concentration via adaptive weighted feature selection and multichannel feature fusion

Surface enhanced Raman spectroscopy (SERS) holds great potential due to its rapid detection and high sensitivity. However, issues such as signal noise, fluctuations, and spectral shifts can negatively impact its performance in detecting ractopamine in pork. Hierarchical Gradient Aware Spectral Network (HGASNet) was proposed to address these issues. The key innovations are the Spectral Gradient Weighted Attention (SGWA) and Multi-Channel Peak Attention Mechanism (MCPAM) modules within HGASNet. The SGWA module dynamically adjusts feature weights, enhancing sensitivity to critical Raman spectral shifts. Meanwhile, MCPAM leverages multi-channel attention to better capture long range dependencies and fuse global and local information. Additionally, HGASNet's hierarchical structure incrementally extracts and integrates features at various levels, enabling the model to focus on global features in the higher layers while preserving fine-grained local details in the lower layers. Experimental results show HGASNet outperforming existing approaches, achieving R2 of 0.9972, RMSRE of 0.0413, RMSLE of 0.0548, sMAPE of 6.47 %, and MAPE of 6.72 %.

Simultaneous quantitative analysis of multiple metabolites using label-free surface-enhanced Raman spectroscopy and explainable deep learning

Metabolites serve as vital biomarkers, reflecting physiological and pathological states and offering insights into disease progression and early detection. This study introduces an advanced analytical technique integrating label-free Surface-Enhanced Raman Spectroscopy (SERS) with deep learning, and leverages SHAP (SHapley Additive exPlanations) to provide a visual interpretative analysis of the predictive rationale of the deep learning model, facilitating simultaneous detection and quantitative analysis of multiple metabolites. Monolayer silver nanoparticle SERS substrates were fabricated via a triple-phase interfacial self-assembly method, which captured complex spectral information of target metabolites in mixed solutions. A custom-built deep neural network model with multi-channel feature extraction was employed to predict the concentrations of uric acid (R2 = 0.976), xanthine (R2 = 0.971), hypoxanthine (R2 = 0.977), and creatinine (R2 = 0.940). The method’s scalability was validated as the performance remained consistent with an increasing number of simultaneous targets. This approach offers a sensitive, cost-effective, and rapid alternative for metabolite analysis, with significant implications for clinical diagnostics and personalized medicine.

Decoding Brain Signals from Rapid-Event EEG for Visual Analysis Using Deep Learning.

The perception and recognition of objects around us empower environmental interaction. Harnessing the brain's signals to achieve this objective has consistently posed difficulties. Researchers are exploring whether the poor accuracy in this field is a result of the design of the temporal stimulation (block versus rapid event) or the inherent complexity of electroencephalogram (EEG) signals. Decoding perceptive signal responses in subjects has become increasingly complex due to high noise levels and the complex nature of brain activities. EEG signals have high temporal resolution and are non-stationary signals, i.e., their mean and variance vary overtime. This study aims to develop a deep learning model for the decoding of subjects' responses to rapid-event visual stimuli and highlights the major factors that contribute to low accuracy in the EEG visual classification task.The proposed multi-class, multi-channel model integrates feature fusion to handle complex, non-stationary signals. This model is applied to the largest publicly available EEG dataset for visual classification consisting of 40 object classes, with 1000 images in each class. Contemporary state-of-the-art studies in this area investigating a large number of object classes have achieved a maximum accuracy of 17.6%. In contrast, our approach, which integrates Multi-Class, Multi-Channel Feature Fusion (MCCFF), achieves a classification accuracy of 33.17% for 40 classes. These results demonstrate the potential of EEG signals in advancing EEG visual classification and offering potential for future applications in visual machine models.

Image tampering detection based on RDS-YOLOv5 feature enhancement transformation

Malicious image tampering has gradually become another way to threaten social stability and personal safety. Timely detection and precise positioning can help reduce the occurrence of risks and improve the overall safety of society. Due to the limitations of highly targeted dataset training and low-level feature extraction efficiency, the generalization and actual performance of the recent tampered detection technology have not yet reached expectations. In this study, we propose a tampered image detection method based on RDS-YOLOv5 feature enhancement transformation. Firstly, a multi-channel feature enhancement fusion algorithm is proposed to enhance the tampering traces in tampered images. Then, an improved deep learning model named RDS-YOLOv5 is proposed for the recognition of tampered images, and a nonlinear loss metric of aspect ratio was introduced into the original SIOU loss function to better optimize the training process of the model. Finally, RDS-YOLOv5 is trained by combining the features of the original image and the enhancement image to improve the robustness of the detection model. A total of 6187 images containing three forms of tampering: splice, remove, and copy-move were used to comprehensively evaluate the proposed algorithm. In ablation test, compared with the original YOLOv5 model, RDS-YOLOv5 achieved a performance improvement of 6.46%, 5.13%, and 3.15% on F1-Score, mAP50 and mAP95, respectively. In comparative experiments, using SRIOU as the loss function significantly improved the model’s ability to search for the real tampered regions by 2.54%. And the RDS-YOLOv5 model trained by the fusion dataset further improved the comprehensive detection performance by about 1%.

Methodology and Experimental Verification for Predicting the Remaining Useful Life of Milling Cutters Based on Hybrid CNN-LSTM-Attention-PSA

In modern manufacturing, the prediction of the remaining useful life (RUL) of computer numerical control (CNC) milling cutters is crucial for improving production efficiency and product quality. This study proposes a hybrid CNN-LSTM-Attention-PSA model that combines convolutional neural networks (CNN), long short-term memory (LSTM) networks, and attention mechanisms to predict the RUL of CNC milling cutters. The model integrates cutting force, vibration, and current signals for multi-channel feature extraction during cutter wear. The model’s hyperparameters are optimized using a PID-based search algorithm (PSA), and comparative experiments were conducted with different predictive models. The experimental results demonstrate the proposed model’s superior performance compared to CNN, LSTM, and hybrid CNN-LSTM models, achieving an R2 score of 99.42% and reducing MAE, RMSE, and MAPE by significant margins. The results validate that the proposed method has significant reference and practical value for RUL prediction research of CNC milling cutters.

Steel Wire Rope Damage Width Identification Method Based on Residual Networks and Multi-Channel Feature Fusion

In order to ensure the safety of steel wire rope in various application scenarios, it is particularly important to quantitatively detect the defects of wire rope. Complex detection conditions affect the detection efficiency of wire rope. Therefore, based on the magnetic flux leakage method, this study proposes a method to identify the damage width of steel wire rope for multi-channel fusion of a Hall sensor array. Firstly, the Hall sensor array is used to capture the magnetic flux leakage data of steel wire rope; then, continuous wavelet transform is used to decompose the original data, and moving average filtering is used to denoise each component; the denoised components are merged and converted into a time spectrum, and the time spectrum is classified by ResNet50 image classification model to realize the detection of wire rope damage width. According to the dataset used in this study, the results show that the proposed method performs best in the mainstream noise reduction model; detection accuracy for the width of damage in steel wire ropes is 97%, which proves that the proposed method is effective and feasible.

ACP-Net: Asymmetric Center Positioning Network for Real-Time Text Detection

Scene text detection is crucial across numerous application fields. However, despite the emphasis on real-time performance in scene text detection, most existing detection models utilize the Feature Pyramid Network (FPN) for feature extraction, often disregarding its inherent limitations. Integrating high-resolution multi-channel features into FPN requires substantial computational resources. While FPN treats local and global features equally and is stable in various applications, its suitability for text-specific features is questionable. To this end, we propose the Asymmetric Center Positioning Network (ACP-Net) to replace FPN, achieving accuracy and real-time text detection in complex scenarios. ACP-Net features an asymmetric feature structure with independent branches for global and local information, along with an adaptive weighted fusion module to capture long-range dependencies effectively. In addition, a text center positioning module enhances text feature understanding by learning feature centers. Comprehensive evaluations across various terminals confirmed ACP-Net’s superior accuracy and speed.

A depth graph attention-based multi-channel transfer learning network for fluid classification from logging data

Fluid classification is a fundamental task in the field of geological sciences to achieve effective reservoir characterization and hydrocarbon exploration. Traditional fluid classification methods are often limited by long processing times and an inability to capture complex relationships within the data. To address this issue, this paper proposes a novel deep learning approach—the Deep Graph Attention Multi-channel Transfer Learning Network (DGMT), aimed at improving the efficiency and accuracy of fluid classification from logging data. This model comprises three key components: a graph attention layer, a multi-channel feature extractor, and a transfer learning module. The graph attention layer is designed to handle spatial dependencies between different logging channels, enhancing classification accuracy by focusing on critical features. The multi-channel feature extractor integrates information from various data sources, ensuring comprehensive utilization of the rich information in logging data. The transfer learning module allows the model to transfer knowledge from pre-trained models of similar tasks, accelerating the training process and significantly improving the model's generalization ability and robustness. This feature enables the DGMT model to adapt to different geological environments and logging conditions, showing superior performance over traditional methods. To validate the effectiveness of the DGMT model, we conducted experiments on actual logging datasets containing multiple oil wells. The experimental results indicate that, compared to common machine learning algorithms and other deep learning methods, the DGMT model significantly improves in accuracy and other classification performance metrics.

Remaining useful life prediction of rolling bearings based on CNN-GRU-MSA with multi-channel feature fusion

ABSTRACT Due to the complex and diverse operating conditions of rolling bearings, it is difficult to accurately predict remaining useful life (RUL) of rolling bearings via traditional prediction models. Besides, when a rolling bearing malfunctions, only single channel or single-domain degradation information is considered to bear RUL prediction, and the prediction effect of existing RUL prediction methods is greatly limited. Therefore, in order to address the above issue, a novel method abbreviated as CNN-GRU-MSA with multi-channel feature fusion is proposed for RUL prediction of rolling bearings. Firstly, the statistical features related to the time series are calculated, and the similarity features are constructed based on the obtained statistical features. Then, the sensitive features are selected and fused through specific indicators. Finally, the dual-channel feature fusion is performed to construct a health indicator (HI), which is input into the proposed CNN-GRU-MSA model for training and achieving RUL prediction of rolling bearings. The effectiveness of the proposed method is validated using two benchmark datasets (i.e. IEEE PHM 2012 challenge datasets and XJTU-SY datasets). Experimental results demonstrate the superiority and effectiveness of the proposed method over other similar prediction methods in bearing RUL prediction.

AID-YOLO: An Efficient and Lightweight Network Method for Small Target Detector in Aerial Images

The progress of object detection technology is crucial for obtaining extensive scene information from aerial perspectives based on computer vision. However, aerial image detection presents many challenges, such as large image background sizes, small object sizes, and dense distributions. This research addresses the specific challenges relating to small object detection in aerial images and proposes an improved YOLOv8s-based detector named Aerial Images Detector-YOLO(AID-YOLO). Specifically, this study adopts the General Efficient Layer Aggregation Network (GELAN) from YOLOv9 as a reference and designs a four-branch skip-layer connection and split operation module Re-parameterization-Net with Cross-Stage Partial CSP and Efficient Layer Aggregation Networks (RepNCSPELAN4) to achieve a lightweight network while capturing richer feature information. To fuse multi-scale features and focus more on the target detection regions, a new multi-channel feature extraction module named Convolutional Block Attention Module with Two Convolutions Efficient Layer Aggregation Net-works (C2FCBAM) is designed in the neck part of the network. In addition, to reduce the sensitivity to position bias of small objects, a new function, Normalized Weighted Distance Complete Intersection over Union (NWD-CIoU_Loss) weight adaptive loss function, was designed in this study. We evaluate the proposed AID-YOLO method through ablation experiments and comparisons with other advanced models on the VEDAI (512, 1024) and DOTAv1.0 datasets. The results show that compared to the Yolov8s baseline model, AID-YOLO improves the mAP@0.5 metric by 7.36% on the VEDAI dataset. Simultaneously, the parameters are reduced by 31.7%, achieving a good balance between accuracy and parameter quantity. The Average Precision (AP) for small objects has improved by 8.9% compared to the baseline model (YOLOv8s), making it one of the top performers among all compared models. Furthermore, the FPS metric is also well-suited for real-time detection in aerial image scenarios. The AID-YOLO method also demonstrates excellent performance on infrared images in the VEDAI1024 (IR) dataset, with a 2.9% improvement in the mAP@0.5 metric. We further validate the superior detection and generalization performance of AID-YOLO in multi-modal and multi-task scenarios through comparisons with other methods on different resolution images, SODA-A and the DOTAv1.0 datasets. In summary, the results of this study confirm that the AID-YOLO method significantly improves model detection performance while maintaining a reduced number of parameters, making it applicable to practical engineering tasks in aerial image object detection.

A soft robotic system imitating the multimodal sensory mechanism of human fingers for intelligent grasping and recognition

The sensory function is crucial for achieving precise feedback control and environmental interaction in soft robots. Therefore, the multimodal sensory capabilities that mimic human fingers have always been a research hotspot. This study proposes a design method for a soft robotic gripper that can emulate the multimodal perception mechanism of human fingers and achieve high-precision recognition of grasped objects using machine learning algorithms. In addition, a finger texture structure inspired by human fingerprints is designed using a negative pressure-induced buckling method to enhance the soft gripper's ability to perceive minute surface textures. Inspired by slow adapting (SA) receptors and fast adapting (FA) receptors of human fingers, we have innovatively designed a capacitive curvature sensor (CS) and a triboelectric texture sensor (TS) to detect the size, material, and texture of objects. Furthermore, a non-contact medium identification sensor (MIS), having a sensitivity of more than 1300 times higher compared to the traditional MISs, is fabricated using the principle of electromagnetic resonance. It is shown that by employing a deep learning-based multi-channel feature fusion network, the soft grasping system with multimodal sensing has achieved a recognition accuracy of 98.43 % for different objects. This work provides an innovative approach for the application of soft robots in various fields, such as logistics sorting and food safety.

A multi-channel fusion variational autoencoder-based RUL prediction approach for multi-sensor systems

Abstract Deep learning (DL)-based approaches have demonstrated remarkable performance in predicting the remaining useful life (RUL) of complex systems, which is beneficial for making timely maintenance decisions. However, the majority of these DL methods suffer from a lack of interpretability, and it is difficult to mine the degradation features in the presence of significant measurement noises. To remedy the deficiency, a multi-channel fusion variational autoencoder (MCFVAE)-based approach is proposed. A feature fusion module is designed to capture and fuse the multi-channel features, which facilitates the disclosure of the degradation information from the multi-sensor data. A variational inference module is further introduced to generate the compressive representations and project them into a latent space as an interpretable component, which can display the degradation degree of the multi-sensor systems. A regressor module is finally utilized to establish the relationship between the compressive representations and the RUL. The superior feature fusion and distribution characteristics learning abilities of the MCFVAE contribute to achieving robust and interpretable RUL prediction. The effectiveness and superiority of the proposed method are experimentally validated through a publicly available Commercial modular aero propulsion system simulation dataset and compared with the existing methods.

MAA-YOLOv8: enhanced steel surface defect detection through multi-head attention mechanism and lightweight feature fusion

In the process of industrial production, product defects often arise due to improper operations among other reasons, rendering the detection of such flaws an indispensable procedure. However, the vast array of defect types, coupled with their complex characteristics, poses ongoing challenges for contemporary defect detection algorithms within industrial settings. To solve this problem, the present study introduces an enhanced steel surface defect detection model based on the modified YOLOv8 algorithm-termed the MAA-YOLOv8 model-to augment the accuracy and practicality of the algorithm. Initially, a multi-head attention mechanism was incorporated into the C2f to bolster the feature extraction capabilities within the backbone network and diversify the attention maps. Secondly, in the neck structure, we design a multi-channel feature fusion module (McPAN) to solve the problem of balance between computational efficiency and the ability to capture useful features. A series of experiments conducted on the NEU-DET dataset reveal that the MAA-YOLOv8 model achieves a mean Average Precision (mAP) of 94.4%, representing an enhancement of 11.1% over the original YOLOv8s model. The MAA-YOLOv8 model proposed in this study substantially elevates the performance of steel surface defect detection while ensuring the speed of detection.

MFS-DBF: A trustworthy multichannel feature sieve and decision boundary formulation system for Obstructive Sleep Apnea detection

The fine identification of sleep apnea events is instrumental in Obstructive Sleep Apnea (OSA) diagnosis. The development of sleep apnea event detection algorithms based on polysomnography is becoming a research hotspot in medical signal processing. In this paper, we propose an Inverse-Projection based Visualization System (IPVS) for sleep apnea event detection algorithms. The IPVS consists of a feature dimensionality reduction module and a feature reconstruction module. First, features of blood oxygen saturation and nasal airflow are extracted and used as input data for event analysis. Then, visual analysis is conducted on the feature distribution for apnea events. Next, dimensionality reduction and reconstruction methods are combined to achieve the dynamic visualization of sleep apnea event feature sets and the visual analysis of classifier decision boundaries. Moreover, the decision-making consistency is explored for various sleep apnea event detection classifiers, which provides researchers and users with an intuitive understanding of the detection algorithm. We applied the IPVS to an OSA detection algorithm with an accuracy of 84% and a diagnostic accuracy of 92% on a publicly available dataset. The experimental results show that the consistency between our visualization results and prior medical knowledge provides strong evidence for the practicality of the proposed system. For clinical practice, the IPVS can guide users to focus on samples with higher uncertainty presented by the OSA detection algorithm, reducing the workload and improving the efficiency of clinical diagnosis, which in turn increases the value of trust.

YOLO-MPAM: Efficient real-time neural networks based on multi-channel feature fusion

Object detection is a crucial foundation in the field of autonomous driving. Therefore, accurate and real-time detection of road objects, including vehicles and pedestrians, is essential for the success of autonomous vehicles. Convolutional neural networks represented by YOLOv8 have made significant progress in autonomous driving. However, there are also various problems, such as false recognition, missed detection, and susceptibility to complex weather. Aiming at the problem of low traffic sign detection effect in automatic driving scenes, an improved YOLOv8 automatic driving target detection algorithm is proposed. Firstly, Mosaic data augmentation and Multi-Path Attention Mechanism (MPAM) are introduced based on the YOLOv8 object detection model. The improved model can obtain location information from the feature map in the early stages of recognition, enabling the model to focus more on the regions of interest. Secondly, because the model pays different attention to global and local features, to make the model pay more attention to fine-grained small targets, so a complex bidirectional fusion structure is proposed in the neural network. Finally, we comprehensively consider the ratio of predicted boxes to ground-truth. The CIoU-tune loss function replaces the original IoU loss function, minimizing the height and width discrepancies between predicted and actual bounding boxes. This improvement accelerates the convergence speed of the model, achieving better localization results. After undergoing ablation analysis, the YOLO-MPAM method is compared with other methods on dataset TT100K. The mAP50 value of the improved algorithm is 86.1%, showing a 9.6% improvement compared to YOLOv8n. Additionally, the model achieves higher accuracy in detecting small objects, effectively enhancing the detection performance of the model at the same inference speed.

Multiple source direction-of-arrival estimation applicable under near-, far-, and mixed-field scenarios.

Traditionally, direction-of-arrival (DOA) estimations under near- and far-field scenarios are treated as independent tasks based on the corresponding acoustic model, hence necessitating a proper soundfield detector as an upstream processing tool, whereas there may not be a distinct boundary between different soundfield types, especially the mixed-field scenarios where both near- and far-field sources coexist simultaneously. To handle this issue, this article investigates a multisource DOA estimator that equally localizes multiple near-, far-, and mixed-field sources, not requiring any specialized adjustments. We (i) define a signal-invariant multichannel feature denoted generalized relative harmonic coefficients in the spherical harmonics domain; (ii) derive the analytical expression of this feature and summarize its unique properties, exhibiting consistence for both near- and far-field sources; (iii) estimate source elevation and azimuth using the magnitude and phase parts of this feature, respectively; (iv) detect single-source dominated periods from the mixed measurements based on an investigated distance measure; and (v) count the number of sources and localize their DOAs by clustering the single-source dominated estimates. Extensive experimental results, in both simulated and real-life environments, finally confirm the effectiveness of the proposed algorithm under diverse acoustic scenarios, and a superiority over baseline approaches in localizing mixed-field sources.

MultiWaveNet: A long time series forecasting framework based on multi-scale analysis and multi-channel feature fusion

Long time series forecasting is widely used in areas such as power dispatch, traffic control, and weather forecasting. The pattern of seasonality and trends in long time series are often complex, especially when they are presented at different time scales. Existing methods typically focus on only one scale or randomly select scales, which leads to a significant loss of valuable information. Additionally, current methods often transform multi-channel data into a single-channel format, ignoring interactions and complex relationships between channels. The paper proposes MultiWaveNet, a novel long time series forecasting framework that addresses seasonality as well as trends separately. For the seasonal component, the framework uses multi-scale wavelet decomposition to generate subseries at multiple scales. A learnable optimization factor is introduced simultaneously to separate high-frequency components mixed in low-frequency series after wavelet decomposition. In order to reduce information redundancy and model complexity, the paper develops a wavelet domain sampling encoder that consists of just one Transformer encoder, ensuring effective modeling of long-term dependencies while maintaining feature extraction effectiveness. As for the trend component, unlike previous research, the weights of channels are adjusted based on their importance, allowing the more crucial channels to have a greater impact and thereby addressing the limitations of individual processing methods. The paper performs extensive experiments on nine standard datasets, demonstrating that MultiWaveNet is the most competitive method.

FCNet: a deep neural network based on multi-channel feature cascading for image denoising

FCNet: a deep neural network based on multi-channel feature cascading for image denoising

Simplified neural architecture for efficient human motion prediction in human-robot interaction

Human motion prediction is essential for safe and effective human-robot interaction, but modeling the intricate spatio-temporal dynamics inherent in human movement remains challenging. While recent methods have advanced motion prediction accuracy, they rely on complex neural architectures such as Graph Convolutional Networks and Recurrent Neural Networks, demanding extensive hyperparameter tuning. To overcome this limitation, we propose a streamlined multi-stage layer incremental perceptron (MSLP) architecture that achieves competitive results with far fewer parameters, improving efficiency. The MSLP eschews the complexity of prevalent networks and instead takes a stepped refinement approach to predict intricate motions using a lightweight model. This simplified yet effective design enables nuanced learning of spatio-temporal relationships without exhaustive tuning. Specifically, the MSLP incorporates two key components: a multi-channel feature extraction and enhancement block (MFE-block) and a temporal feature extraction module (TFE-block). The MFE-block strengthens the representation of each action node by integrating multi-dimensional action features. The TFE-block then captures the contextual relationships between actions over time. Together, the MFE-block and TFE-block allow the MSLP to model the complex spatial and temporal dynamics of human movement using a streamlined architecture and minimal parameter tuning. When evaluated on established datasets including Human3.6 M, CMU, 3DPW, and AMASS, the proposed MSLP method achieves improved accuracy gains of 16.7%-67.0% over existing state-of-the-art techniques. Additionally, the MSLP significantly reduces the number of parameters by 35.7%-99.5% compared to prior architectures. We find that the proposed MSLP significantly improves both long-term prediction and generalization capabilities of the model.

Fault Diagnosis of Gear Based on Multichannel Feature Fusion and DropKey-Vision Transformer

Fault Diagnosis of Gear Based on Multichannel Feature Fusion and DropKey-Vision Transformer