One-dimensional Convolutional Neural Network Research Articles

Fault Diagnosis Method for Tractor Transmission System Based on Improved Convolutional Neural Network–Bidirectional Long Short-Term Memory

In response to the problems of limited algorithms and low diagnostic accuracy for fault diagnosis in large tractor transmission systems, as well as the high noise levels in tractor working environments, a defect detection approach for tractor transmission systems is proposed using an enhanced convolutional neural network (CNN) and a bidirectional long short-term memory neural network (BILSTM). This approach uses a one-dimensional convolutional neural network (1DCNN) to create three feature extractors of varying scales, directly extracting feature information from different levels of the raw vibration signals. Simultaneously, in order to enhance the model’s predicted accuracy and learn the data features more effectively, it presents the multi-head attention mechanism (MHA). To overcome the issue of high noise levels in tractor working environments and enhance the model’s robustness, an adaptive soft threshold is introduced. Finally, to recognize and classify faults, the fused feature data are fed into a classifier made up of bidirectional long short-term memory (BILSTM) and fully linked layers. The analytical findings demonstrate that the fault recognition accuracy of the method described in this article is over 98%, and it also has better performance in noisy environments.

SOH estimation of lithium-ion batteries based on multi-feature deep fusion and XGBoost

State of Health (SOH) is a crucial metric for battery management systems, and accurate estimation of battery SOH is essential for the underlying management and maintenance of batteries. Traditional data-driven approaches lack in-depth analysis of health features and fail to exploit the deep-level health information among these features. Therefore, this paper extracts six health features from both measurement and calculation perspectives based on actual measurable charging curve data, and conducts aging analysis of health features from a mechanistic perspective and Spearman correlation analysis from a quantitative perspective. Additionally, a strategy combining deep learning with ensemble learning is proposed. This strategy utilizes one-dimensional convolutional neural networks to deeply integrate multiple original features, extracting fusion features with deep-level health information. Subsequently, XGBoost is employed to learn the fusion features for estimating battery SOH. Finally, the effectiveness and superiority of the proposed method are verified on battery aging datasets from the University of Maryland and our laboratory. Experimental results demonstrate that the RMSE and MAE of the proposed method are both below 1 % and outperform other methods.

Predicting Rail Corrugation Based on Convolutional Neural Networks Using Vehicle’s Acceleration Measurements

This paper presents a deep learning approach for predicting rail corrugation based on on-board rolling-stock vertical acceleration and forward velocity measurements using One-Dimensional Convolutional Neural Networks (CNN-1D). The model’s performance is examined in a 1:10 scale railway system at two different forward velocities. During both the training and test stages, the CNN-1D produced results with mean absolute percentage errors of less than 5% for both forward velocities, confirming its ability to reproduce the corrugation profile based on real-time acceleration and forward velocity measurements. Moreover, by using a Gradient-weighted Class Activation Mapping (Grad-CAM) technique, it is shown that the CNN-1D can distinguish various regions, including the transition from damaged to undamaged regions and one-sided or two-sided corrugated regions, while predicting corrugation. In summary, the results of this study reveal the potential of data-driven techniques such as CNN-1D in predicting rails’ corrugation using online data from the dynamics of the rolling-stock, which can lead to more reliable and efficient maintenance and repair of railways.

Blood CO Status Classification Using UV-VIS Spectroscopy and PSO-optimized 1D-CNN Model

Rapid and effective blood carbon monoxide (CO) assessment is of great importance, especially in estimating CO-related morbidity and instituting effective preventive measures. The conventional detection methods using CO breath analysis lack sensitivity, while collecting biological fluid samples for CO level measurement is prone to external contamination and expensive for frequent use. This study proposes a one-dimensional convolutional neural network (1D-CNN) consisting of three stacked biconvolutional layers for binary classification of blood CO status using the diffuse reflectance spectroscopy technique. Iterative particle swarm optimization (PSO) has efficiently found the best network parameters to learn important features from the reflectance spectroscopy data. The findings showed good testing accuracy, specificity, and precision of 92.9%, 90%, and 89.7%, respectively, and a high sensitivity of 96.3% in determining abnormal blood CO among smokers using the proposed CNN network. Comparisons with eight existing machine learning and deep learning models revealed the proposed method’s effectiveness in classifying blood CO status while reducing computing time by 8–13 folds. The findings of this work provide new insights that are valuable for researchers in neural network design automation, healthcare management, and skin-related research, specifically for application in nondestructive evaluation and clinical decision-making.

Prediction of Sunspot Number with Hybrid Model Based on 1D-CNN, BiLSTM and Multi-Head Attention Mechanism

Sunspots have a significant impact on human activities. In this study, we aimed to improve solar activity prediction accuracy. To predict the sunspot number based on different aspects, such as extracted features and relationships among data, we developed a hybrid model that includes a one-dimensional convolutional neural network (1D-CNN) for extracting the features of sunspots and bidirectional long short-term memory (BiLSTM) embedded with a multi-head attention mechanism (MHAM) to learn the inner relationships among data and finally predict the sunspot number. We evaluated our model and several existing models according to different evaluation indicators, such as mean absolute error (MAE) and root mean square error (RMSE). Compared with the informer, stacked LSTM, XGBoost-DL, and EMD-LSTM-AM models, the RMSE and MAE of our results were more than 42.5% and 65.1% lower, respectively. The experimental results demonstrate that our model has higher accuracy than other methods.

Application of Attention-Enhanced 1D-CNN Algorithm in Hyperspectral Image and Spectral Fusion Detection of Moisture Content in Orah Mandarin (Citrus reticulata Blanco)

A method fusing spectral and image information with a one-dimensional convolutional neural network(1D-CNN) for the detection of moisture content in Orah mandarin (Citrus reticulata Blanco) was proposed. The 1D-CNN model integrated with three different attention modules (SEAM, ECAM, CBAM) and machine learning models were applied to individual spectrum and fused information by passing the traditional feature extraction stage. Additionally, the dimensionality reduction of hyperspectral images and extraction of one-dimensional color and textural features from the reduced images were performed, thus avoiding the large parameter volumes and efficiency decline inherent in the direct modeling of two-dimensional images. The results indicated that the 1D-CNN model with integrated attention modules exhibited clear advantages over machine learning models in handling multi-source information. The optimal machine learning model was determined to be the random forest (RF) model under the fusion information, with a correlation coefficient (R) of 0.8770 and a root mean square error (RMSE) of 0.0188 on the prediction set. The CBAM-1D-CNN model under the fusion information exhibited the best performance, with an R of 0.9172 and an RMSE of 0.0149 on the prediction set. The 1D-CNN models utilizing fusion information exhibited superior performance compared to single spectrum, and 1D-CNN with the fused information based on SEAM, ECAM, and CBAM respectively improved Rp by 4.54%, 0.18%, and 10.19% compared to the spectrum, with the RMSEP decreased by 11.70%, 14.06%, and 31.02%, respectively. The proposed approach of 1D-CNN integrated attention can obtain excellent regression results by only using one-dimensional data and without feature pre-extracting, reducing the complexity of the models, simplifying the calculation process, and rendering it a promising practical application.

Moving object detection via feature extraction and classification

Abstract Foreground segmentation (FS) plays a fundamental and important role in computer vision, but it remains a challenging task in dynamic backgrounds. The supervised method has achieved good results, but the generalization ability needs to be improved. To address this challenge and improve the performance of FS in dynamic scenarios, a simple yet effective method has been proposed that leverages superpixel features and a one-dimensional convolution neural network (1D-CNN) named SPF-CNN. SPF-CNN involves several steps. First, the coined Iterated Robust CUR (IRCUR) is utilized to obtain candidate foregrounds for an image sequence. Simultaneously, the image sequence is segmented using simple linear iterative clustering. Next, the proposed feature extraction approach is applied to the candidate matrix region corresponding to the superpixel block. Finally, the 1D-CNN is trained using the obtained superpixel features. Experimental results demonstrate the effectiveness of SPF-CNN, which also exhibits strong generalization capabilities. The average F1-score reaches 0.83.

Online meta-learning approach for sensor fault diagnosis using limited data

The accurate and timely diagnosis of sensor faults plays a critical role in ensuring the reliability and performance of structural health monitoring (SHM) systems. However, the challenge is detecting, locating, and estimating sensor faults in an online manner using limited training data. To resolve this problem, a novel approach for online sensor fault diagnosis is proposed for SHM. The proposed approach is based on meta-learning, which enables superior model generalization capabilities using limited data. The detection, localization, and estimation of typical sensor faults in an online manner can be achieved efficiently by the proposed approach. First, a one-dimensional convolutional neural network (1D CNN) is designed to detect and locate faulty sensors. The initial model parameters of the 1D CNN are optimized using a model-agnostic meta-learning training strategy. This strategy allows the acquisition of transferable prior knowledge, which can speed up the learning process on new sensor fault detection and localization tasks. The meta-learning strategy also enables efficient and accurate detection and localization of potential faulty sensors with limited data. After detecting and locating the faulty sensors, an online updating algorithm based on a dual Kalman filter is used to estimate the severity of sensor faults and structural states simultaneously. The proposed approach is demonstrated with simulated sensor faults that cover a numerical example and field measurements from the Canton Tower. The results show that the proposed approach is applicable for online sensor fault diagnosis in SHM.

Moving object detection via feature extraction and classification

Abstract Foreground segmentation (FS) plays a fundamental and important role in computer vision, but it remains a challenging task in dynamic backgrounds. The supervised method has achieved good results, but the generalization ability needs to be improved. To address this challenge and improve the performance of FS in dynamic scenarios, a simple yet effective method has been proposed that leverages superpixel features and a one-dimensional convolution neural network (1D-CNN) named SPF-CNN. SPF-CNN involves several steps. First, the coined Iterated Robust CUR (IRCUR) is utilized to obtain candidate foregrounds for an image sequence. Simultaneously, the image sequence is segmented using simple linear iterative clustering. Next, the proposed feature extraction approach is applied to the candidate matrix region corresponding to the superpixel block. Finally, the 1D-CNN is trained using the obtained superpixel features. Experimental results demonstrate the effectiveness of SPF-CNN, which also exhibits strong generalization capabilities. The average F1-score reaches 0.83.

Research on CO concentration detection based on deep learning and TDLAS technology

In order to improve the accuracy of detecting CO and other gases in the industrial field, a CO concentration detection scheme based on tunable diode laser absorption spectroscopy (TDLAS) and the CO absorption spectral line at 1566 nm was proposed. At the outset, after removing the outliers from the original data, the harmonic signal is filtered using a moving average filtering method to enhance the signal-to-noise ratio. Subsequently, a one-dimensional convolutional neural network (1D-CNN) is employed to extract the peak and valley values, as well as the mean values of each segment of harmonic data. In the end, the Long short term memory network (LSTM) model establishes a nonlinear fitting relationship between gas characteristics and concentration, and the attention layer is added to further optimize the network weight parameters. The experimental results indicate that the R2 for the test set is 0.9884, and the RMSE is 156.869. Through Allan deviation analysis, the detection limit reaches 1.9605 ppm at an integration time of 60 s. Compared to the traditional regression model, the accuracy and stability are significantly improved, making it highly suitable for gas detection in the industrial circle.

Prediction of Aerodynamic Forces at the Tip of the Compressor Blades Based on Multi-scale 1DCNN Combined with CBAM

The compressor is a crucial component of aircraft engines, and the blades are the critical factor affecting the performance of the compressor. Based on multi-scale one-dimensional convolution neural network (1DCNN) with Convolutional Block Attention Module (CBAM), a data-driven model is proposed for predicting the aerodynamic characteristics of the blade tips. The model is trained using the Adam with decoupled weight decay (AdamW) optimizer and a staged learning rate scheduling strategy. Due to the distinct aerodynamic pressure distributions on the suction and pressure sides, separate models are constructed in order to reveal the aerodynamic performance of the blade tips accurately. During the model validation, Root Mean Square Error (RMSE) and the coefficient of determination (R2) are taking as evaluation criterions, where high reliability is demonstrated compared to Computational Fluid Dynamics (CFD) results.

Deepvirusclassifier: a deep learning tool for classifying SARS-CoV-2 based on viral subtypes within the coronaviridae family

PurposeIn this study, we present DeepVirusClassifier, a tool capable of accurately classifying Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) viral sequences among other subtypes of the coronaviridae family. This classification is achieved through a deep neural network model that relies on convolutional neural networks (CNNs). Since viruses within the same family share similar genetic and structural characteristics, the classification process becomes more challenging, necessitating more robust models. With the rapid evolution of viral genomes and the increasing need for timely classification, we aimed to provide a robust and efficient tool that could increase the accuracy of viral identification and classification processes. Contribute to advancing research in viral genomics and assist in surveilling emerging viral strains.MethodsBased on a one-dimensional deep CNN, the proposed tool is capable of training and testing on the Coronaviridae family, including SARS-CoV-2. Our model’s performance was assessed using various metrics, including F1-score and AUROC. Additionally, artificial mutation tests were conducted to evaluate the model’s generalization ability across sequence variations. We also used the BLAST algorithm and conducted comprehensive processing time analyses for comparison.ResultsDeepVirusClassifier demonstrated exceptional performance across several evaluation metrics in the training and testing phases. Indicating its robust learning capacity. Notably, during testing on more than 10,000 viral sequences, the model exhibited a more than 99% sensitivity for sequences with fewer than 2000 mutations. The tool achieves superior accuracy and significantly reduced processing times compared to the Basic Local Alignment Search Tool algorithm. Furthermore, the results appear more reliable than the work discussed in the text, indicating that the tool has great potential to revolutionize viral genomic research.ConclusionDeepVirusClassifier is a powerful tool for accurately classifying viral sequences, specifically focusing on SARS-CoV-2 and other subtypes within the Coronaviridae family. The superiority of our model becomes evident through rigorous evaluation and comparison with existing methods. Introducing artificial mutations into the sequences demonstrates the tool’s ability to identify variations and significantly contributes to viral classification and genomic research. As viral surveillance becomes increasingly critical, our model holds promise in aiding rapid and accurate identification of emerging viral strains.

Chemical composition prediction in goji (Lycium barbarum) using hyperspectral imaging and multi-task 1DCNN with attention mechanism

The bioactive components of goji berries (Lycium barbarum) are crucial determinants of their nutritional and commercial value. In this study, we combined hyperspectral imaging technology (HSI) with a one-dimensional convolutional neural network (1DCNN) to predict the content of chemical compositions in goji. To enhance the model's ability to focus on relevant information, we introduced the channel attention module (CAM), spectral attention module (SAM) and fused them together, which can focus on output features of the convolution kernels differently and adaptively emphasize more effective spectral bands, respectively. Moreover, considering the limitations of traditional single-task prediction methods and the inherent correlations among different constituents, we employed a multi-task CNN for the simultaneous prediction of various goji constituents. The results indicate that the attention-enhanced 1DCNN model outperforms both the partial least squares regression (PLSR) model and the vanilla 1DCNN. With multi-task learning, the model achieves optimal performance, achieving an average R2 of 0.9435 for the prediction of the three components. Our research develops an efficient and accurate method for predicting the constituents of goji berries, providing a valuable, convenient, and effective tool for assessing and detecting the quality of fruit products in the food industry.

A novel one-dimensional convolutional neural network with parallel attention for fault diagnosis of rigid guides

A novel one-dimensional convolutional neural network with parallel attention for fault diagnosis of rigid guides

Quality assessment of Gastrodia elata Blume (Tianma) based on Vis-NIR spectroscopy: Discrimination of harvest times and prediction of quality indicator contents

Gastrodia elata Blume (Tianma) is widely cultivated and consumed as a food and herbal material in China. With its increasing use as a functional food, Tianma is gaining popularity in the market. However, during the cultivation process, the harvest time affects its maturity and storage time, while the total content of gastrodin (GAS) and p-hydroxybenzyl alcohol (HA) affects its nutritional value, making it difficult to control the quality of Tianma. In this study, a combination of visible-near-infrared (Vis-NIR) spectroscopy and chemometrics was used to identify the harvest time and determine the quality indicator content. A Vis-NIRRamanNet modeling method based on the architecture of RamanNet was proposed and compared with harvest time discrimination models constructed using partial least squares discriminant analysis (PLS-DA), k-nearest neighbor (KNN), support vector machine (SVM), and one-dimensional convolutional neural network (1D-CNN). The results demonstrated that Vis-NIRRamanNet exhibited superior performance, achieving accuracy rates of 0.995 and 0.973 on the training and testing sets, respectively. Subsequently, Vis-NIR spectroscopy combined with support vector regression (SVR) was employed for quantitative analysis of quality indicator content. Wavelength selection methods such as uninformative variable elimination (UVE), genetic algorithm (GA), and competitive adaptive reweighted sampling (CARS) were utilized to optimize the model. The results indicated that the CARS-SVR model provided the best prediction of quality indicator content, with root mean square error of prediction (RMSEP) and coefficient of determination (Rp2) values of 0.257 and 0.901, respectively. Therefore, the combination of Vis-NIR spectroscopy and chemometrics shows significant potential in assisting the selection of high-quality Tianma.

Low Tropospheric Wind Forecasts in Aviation: The Potential of Deep Learning for Terminal Aerodrome Forecast Bulletins

In aviation, accurate wind prediction is crucial, especially during takeoff and landing at complex sites like Gran Canaria Airport. This study evaluated five Deep Learning models: Long Short-Term Memory (LSTM), Vanilla Recurrent Neural Network (vRNN), One-Dimensional Convolutional Neural Network (1dCNN), Convolutional Neural Network Long Short-Term Memory (CNN-LSTM), and Gated Recurrent Unit (GRU) for forecasting wind speed and direction. The LSTM model demonstrated the highest precision, particularly for extended forecasting periods, achieving a mean absolute error (MAE) of 1.23 m/s and a circular MAE (cMAE) of 15.80° for wind speed and direction, respectively, aligning with World Meteorological Organization standards for Terminal Aerodrome Forecasts (TAF). While the GRU and CNN-LSTM also showed promising results, and the 1dCNN excelled in wind direction forecasting over shorter intervals, the vRNN lagged in performance. Additionally, the autoregressive integrated moving average model underperformed relative to the DL models, underscoring the potential of DL, particularly LSTM, in enhancing TAF accuracy at airports with intricate wind patterns. This study not only confirms the superiority of DL over traditional methods but also highlights the promise of integrating artificial intelligence into TAF automation.

Improving skin cancer detection by Raman spectroscopy using convolutional neural networks and data augmentation.

Our previous studies have demonstrated that Raman spectroscopy could be used for skin cancer detection with good sensitivity and specificity. The objective of this study is to determine if skin cancer detection can be further improved by combining deep neural networks and Raman spectroscopy. Raman spectra of 731 skin lesions were included in this study, containing 340 cancerous and precancerous lesions (melanoma, basal cell carcinoma, squamous cell carcinoma and actinic keratosis) and 391 benign lesions (melanocytic nevus and seborrheic keratosis). One-dimensional convolutional neural networks (1D-CNN) were developed for Raman spectral classification. The stratified samples were divided randomly into training (70%), validation (10%) and test set (20%), and were repeated 56 times using parallel computing. Different data augmentation strategies were implemented for the training dataset, including added random noise, spectral shift, spectral combination and artificially synthesized Raman spectra using one-dimensional generative adversarial networks (1D-GAN). The area under the receiver operating characteristic curve (ROC AUC) was used as a measure of the diagnostic performance. Conventional machine learning approaches, including partial least squares for discriminant analysis (PLS-DA), principal component and linear discriminant analysis (PC-LDA), support vector machine (SVM), and logistic regression (LR) were evaluated for comparison with the same data splitting scheme as the 1D-CNN. The ROC AUC of the test dataset based on the original training spectra were 0.886±0.022 (1D-CNN), 0.870±0.028 (PLS-DA), 0.875±0.033 (PC-LDA), 0.864±0.027 (SVM), and 0.525±0.045 (LR), which were improved to 0.909±0.021 (1D-CNN), 0.899±0.022 (PLS-DA), 0.895±0.022 (PC-LDA), 0.901±0.020 (SVM), and 0.897±0.021 (LR) respectively after augmentation of the training dataset (p<0.0001, Wilcoxon test). Paired analyses of 1D-CNN with conventional machine learning approaches showed that 1D-CNN had a 1-3% improvement (p<0.001, Wilcoxon test). Data augmentation not only improved the performance of both deep neural networks and conventional machine learning techniques by 2-4%, but also improved the performance of the models on spectra with higher noise or spectral shifting. Convolutional neural networks slightly outperformed conventional machine learning approaches for skin cancer detection by Raman spectroscopy.

Stability evaluation of open-pit mine slope based on Bayesian optimization 1D-CNN

As mechanized open-pit coal mining intensifies, assessing and predicting slope stability has become increasingly important. To address the limitations of traditional mechanical calculations, numerical simulations, and physical experiments, this paper identifies the key factors impacting slope stability in open-pit mines and develops a multi-parameter sample data set. The study employs hyperparameters optimized using a Bayesian algorithm, introduces additional convolutional layers, and combines the Adam optimizer with dropout techniques to enhance the feature extraction and performance of one-dimensional convolutional neural networks (1D-CNN). This leads to a Bayesian-optimized one-dimensional convolutional neural network (B-1D MCNN) model for predicting slope stability.The study evaluates the classification performance and accuracy of various models for slope stability, including BP neural networks, genetic algorithm-optimized convolutional neural networks, 1D-CNN, and B-1D MCNN, using accuracy, precision, and F1-score as metrics. The analysis also examines the influence of factor indicators and training set length on the model's output to assess its generalization capabilities.The research findings suggest that: (1) the B-1D MCNN model for evaluating slope stability demonstrates the capability to accurately depict the nonlinear correlation between influencing factors and slope stability. (2) Compared with other models, the B-1D MCNN model has shown enhancements of 10.96% to 27.85%, 10.26% to 28.55%, and 8.98% to 25.05% in terms of Accuracy, F1-Score, and Precision, respectively. (3) As the length of the training dataset increases, the performance of the model improves accordingly. (4) The B-1D MCNN model shows a generalization power of 87.5%.

Using gross primary production data and deep transfer learning for crop yield prediction in the US Corn Belt

Estimating crop yield prior to harvest plays a crucial role in agricultural decision making. Machine learning models that use remote sensing (RS) data can provide quick and early estimates of crop yield. Previously, timeseries gross primary production (GPP) synthesized from RS data has been employed in agricultural applications. However, relevant research on using GPP data and deep learning models in crop yield prediction is scarce. The aim of this study was to assess the capability of MODIS-derived GPP data and deep transfer learning to predict crop yield. We developed and trained deep neural network (DNN), one-dimensional convolutional neural network (1D-CNN), and gated recurrent unit (GRU) models to predict county-level corn and soybean yields. Then we evaluated the performance of crop-type and location transfer learning strategies in crop yield prediction. The results of this study can be summarized as follows: (1) GPP data and deep transfer learning have the potential to predict crop yield with reasonable prediction accuracy; (2) in the context of transfer learning, crop-type transfer learning produced a coefficient of determination (R2) ranging from 0.521 to 0.784 for soybean and from 0.644 to 0.903 for corn. Meanwhile, location transfer learning yielded R2 values ranging from 0.690 to 0.931 for the Great Plains (GP) climatic zone and 0.480 to 0.801 for the Eastern Temperate Forest (ETF) climatic zone; and (3) For corn and soybean yield predictions, the 1D-CNN model (R2 = 0.864 for corn and 0.750 for soybean) outperformed the DNN and GRU models. Our results reveal that MODIS-derived GPP data and deep transfer learning can be used to effectively predict county-level crop yield.

Dual Stream Long Short-Term Memory Feature Fusion Classifier for Surface Electromyography Gesture Recognition.

Gesture recognition using electromyography (EMG) signals has prevailed recently in the field of human-computer interactions for controlling intelligent prosthetics. Currently, machine learning and deep learning are the two most commonly employed methods for classifying hand gestures. Despite traditional machine learning methods already achieving impressive performance, it is still a huge amount of work to carry out feature extraction manually. The existing deep learning methods utilize complex neural network architectures to achieve higher accuracy, which will suffer from overfitting, insufficient adaptability, and low recognition accuracy. To improve the existing phenomenon, a novel lightweight model named dual stream LSTM feature fusion classifier is proposed based on the concatenation of five time-domain features of EMG signals and raw data, which are both processed with one-dimensional convolutional neural networks and LSTM layers to carry out the classification. The proposed method can effectively capture global features of EMG signals using a simple architecture, which means less computational cost. An experiment is conducted on a public DB1 dataset with 52 gestures, and each of the 27 subjects repeats every gesture 10 times. The accuracy rate achieved by the model is 89.66%, which is comparable to that achieved by more complex deep learning neural networks, and the inference time for each gesture is 87.6 ms, which can also be implied in a real-time control system. The proposed model is validated using a subject-wise experiment on 10 out of the 40 subjects in the DB2 dataset, achieving a mean accuracy of 91.74%. This is illustrated by its ability to fuse time-domain features and raw data to extract more effective information from the sEMG signal and select an appropriate, efficient, lightweight network to enhance the recognition results.