Search

Intense wind shear (I-WS) near airport runways presents a critical challenge to aviation safety, necessitating accurate and timely classification to mitigate risks during takeoff and landing. This study proposes the application of advanced Residual Network (ResNet) architectures including ResNet34 and ResNet50 for classifying I-WS and Non-Intense Wind Shear (NI-WS) events using Doppler Light Detection and Ranging (LiDAR) data from Hong Kong International Airport (HKIA). Unlike conventional models such as feedforward neural networks (FNNs), convolutional neural networks (CNNs), and recurrent neural networks (RNNs), ResNet provides a distinct advantage in addressing key challenges such as capturing intricate WS dynamics, mitigating vanishing gradient issues in deep architectures, and effectively handling class imbalance when combined with Synthetic Minority Oversampling Technique (SMOTE). The analysis results revealed that ResNet34 outperforms other models with a Balanced Accuracy of 0.7106, Probability of Detection of 0.8271, False Alarm Rate of 0.328, F1-score of 0.7413, Matthews Correlation Coefficient of 0.433, and Geometric Mean of 0.701, demonstrating its effectiveness in classifying I-WS events. The findings of this study not only establish ResNet as a valuable tool in the domain of WS classification but also provide a reliable framework for enhancing operational safety at airports. [ABSTRACT FROM AUTHOR]

Abstract The Internet of Things (IoT) has recently attracted substantial interest because of its diverse applications. In the agriculture sector, automated methods for detecting plant diseases offer numerous advantages over traditional methods. In the current study, a new model is developed to categorize plant diseases within an IoT network. The IoT network is simulated for monitoring crop diseases. Routing is performed with Henry Gas Chicken Swarm Optimization (HGCSO), which is designed by integrating Henry Gas Solubility Optimization (HGSO) and Chicken Swarm Optimization (CSO). The fitness parameters of the model include delay, energy, distance, and link lifetime (LLT). At the Base Station (BS), plant disease categorization is performed by collecting plant leaf images. Preprocessing is done on the input images using median filtering. Various features, such as Histogram of Oriented Gradient (HoG), statistical features, Spider Local Image Features (SLIF), and Local Ternary Patterns (LTP) are extracted. Plant disease categorization is carried out using a Deep Residual Network (DRN), which is trained using the developed Caviar Henry Gas Chicken Swarm Optimization (CHGCSO) that combines the CAViaR model with HGCSO. Comparative results show an accuracy of 94.3%, a maximum sensitivity of 93.3%, a maximum specificity of 92%, and an F1-score of 93%, indicating that the CHGCSO-based DRN outperforms existing methods. Graphic Abstract [ABSTRACT FROM AUTHOR]

Multivariate time series data classification has a wide range of applications in reality. With rapid development of deep learning, convolutional networks are widely used in this task and have achieved the current best performance. However, due to high difficulty and cost of collecting this type of data, labeled data is still scarce. In some tasks, the model shows overfitting, resulting in relatively poor classification performance. In order to improve the classification performance under such situation, we proposed a novel classification method based on transfer learning—similar domain knowledge transfer (call SDKT for short). Firstly, we designed a multivariate time series domain distance calculation method (call MTSDDC for short), which helped selecting the source domain that is most similar to target domain; Secondly, we used ResNet as a pre‐trained classifier, transferred the parameters of the similar domain network to the target domain network and continue to fine‐tune the parameters. To verify our method, we conducted experiments on several public datasets. Our study has also shown that the transfer effect from the source domain to the target domain is highly negatively correlated with the distance between them, with an average Pearson coefficient of −0.78. For the transfer of most similar source domain, compared to the ResNet model without transfer and the current best model, the average accuracy improvements on the datasets we used are 4.01% and 1.46% respectively. [ABSTRACT FROM AUTHOR]

Human activity recognition (HAR) in real-world settings has gained significance due to the growth of Internet of Things (IoT) devices such as smartphones and smartwatches. Nonetheless, limitations such as fluctuating environmental conditions and intricate behavioral patterns have impacted the accuracy of the current procedures. This research introduces an innovative methodology employing a modified deep residual network, called 1D-ResNeXt, for IoT-enabled HAR in uncontrolled environments. We developed a comprehensive network that utilizes feature fusion and a multi-kernel block approach. The residual connections and the split–transform–merge technique mitigate the accuracy degradation and reduce the parameter number. We assessed our suggested model on three available datasets, mHealth, MotionSense, and Wild-SHARD, utilizing accuracy metrics, cross-entropy loss, and F1 score. The findings indicated substantial enhancements in proficiency in recognition, attaining 99.97% on mHealth, 98.77% on MotionSense, and 97.59% on Wild-SHARD, surpassing contemporary methodologies. Significantly, our model attained these outcomes with considerably fewer parameters (24,130–26,118) than other models, several of which exceeded 700,000 parameters. The 1D-ResNeXt model demonstrated outstanding effectiveness under various ambient circumstances, tackling a significant obstacle in practical HAR applications. The findings indicate that our modified deep residual network presents a viable approach for improving the dependability and usability of IoT-based HAR systems in dynamic, uncontrolled situations while preserving the computational effectiveness essential for IoT devices. The results significantly impact multiple sectors, including healthcare surveillance, intelligent residences, and customized assistive devices. [ABSTRACT FROM AUTHOR]

This study aims to capture subtle changes in the pupil, identify relatively weak inter-class changes, extract more abstract and discriminative pupil features, and study a pupil refinement recognition method based on attention mechanisms. Based on the deep learning framework and the ResNet101 deep residual network as the backbone network, a pupil refinement recognition model is established. Among them, the image preprocessing module is used to preprocess the pupil images captured by infrared spectroscopy, removing internal noise from the pupil images. By using the ResNet101 backbone network, subtle changes in the pupil are captured, weak inter-class changes are identified, and different features of the pupil image are extracted. The channel attention module is used to screen pupil features and obtain key pupil features. External attention modules are used to enhance the expression of key pupil feature information and extract more abstract and discriminative pupil features. The Softmax classifier is used to process the pupil features captured by infrared spectra and output refined pupil recognition results. Experimental results show that this method can effectively preprocess pupil images captured by infrared spectroscopy and extract pupil features. This method can effectively achieve fine pupil recognition, and the fine recognition effect is relatively good. [ABSTRACT FROM AUTHOR]

In order to eliminate the impact of the imbalance of the sample size of bonafide speech and fake speech in the training dataset on the performance of synthetic speech detection system and further improve the accuracy of synthetic speech detection, a method of synthetic speech detection was proposed based on self-supervised contrastive learning. In this method, the samples after pitch transformation were regarded as negative samples, and the neural network was trained to make the anchor sample features different from the negative sample features, so that the network could extract the features sensitive to pitch transformation. And the deep residual network was used as the back-end classifier to judge the authenticity of the speech. Experimental results show that, compared with the traditional hand-crafted acoustic features, the deep learning-based and the end-to-end spoofing speech detection systems, the proposed method significantly reduces the equal error rate of the system. The synthetic forged speech detection method based on self-supervised contrastive learning can train the network to extract features sensitive to pitch transformation and will not affect the accuracy of synthetic speech detection because of the imbalance of bonafide and fake speech in the dataset, so the accuracy of synthetic forged speech detection is significantly improved.

Copyright of Telecommunication Engineering is the property of Telecommunication Engineering and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Copyright of Telecommunication Engineering is the property of Telecommunication Engineering and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

A modified ResNet road state recognition algorithm was proposed to address the problems of low accuracy in traditional methods for road state recognition and the complex process of manually extracting road image features, which can recognize 7 types of road states. Firstly, a Dense Block dense connection module is added to the ResNet classic model algorithm to extract more shallow edge features of the road surface state. Then, a dual attention mechanism was introduced in the ResNet residual module to adaptively extract deep key features according to the importance of different channel features. Finally, by adding two fully connected layers and introducing Dropout to suppress overfitting, the recognition of road surface states was ultimately achieved. The experimental results show that the accuracy, precision, recall, F1 score and specificity of the proposed method all reach over 99. 0% . It can effectively extract key feature information and has strong robustness, effectively improving the recognition accuracy of road surface states. [ABSTRACT FROM AUTHOR]

Currently, 3D face modeling has been more and more widely used in medical, animation, games, and other fields. With the application of deep neural network, 3D face modeling method based on deep learning is the main research direction at present. However, the accuracy of the 3D face reconstruction model is generally insufficient. To solve this problem, an improved algorithm model based on deep residual network is proposed in this paper. A new convolutional module of Squeeze-and-Excitation is introduced to increase the channel dependence of the entire model, and then Blendshape is combined to drive the entire face data. The normalization error of the new improved neural network algorithm is only 3.12% compared with other traditional algorithm models. The loss function was reduced by 56.72%, 48.95%, and 41.86%, respectively. Compared with the traditional algorithm, the improved algorithm has higher precision and more perfect expression driving ability. This study contributes to improving the accuracy of the whole 3D face reconstruction technology. [ABSTRACT FROM AUTHOR]

Copyright of Telecommunications Science is the property of Beijing Xintong Media Co., Ltd. and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Diabetes mellitus is a severe condition that has the potential to impair strength. The disease known as diabetes mellitus, which is a chronic condition, is brought on by a significant rise in blood glucose levels. The diagnosis of this condition is made using a variety of chemical and physical testing. Diabetes, however, can harm the organs if it goes undetected. This study develops a hybrid deep-learning technique to recognize Type 2 diabetes mellitus. The data is cleaned up at the pre-processing stage using a data transformation technique based on the Yeo-Jhonson transformation. The tanimoto similarity is used in the feature selection process to select the best features from the data. To prepare data for future processing, data augmentation is performed. The Deep Residual Network and the Rider-based Neural Network are recommended and trained separately for the T2DM identification using the Competitive Multi-Verse Rider Optimizer. The outputs generated by the RideNN and DRN classifiers are blended using correlation-based fusion. The suggested CMVRO-based NN-DRN has shown improved performance with the highest accuracy of 91.4%, sensitivity of 94.8%, and specificity of 90.1%. [ABSTRACT FROM AUTHOR]

Thermal error is a critical factor influencing the machining accuracy of CNC machine tools, so it is essential to comprehensively model and compensate for thermal errors in CNC machine tools. This paper proposes a deep attentional residual network thermal error prediction model driven by thermal image inputs. In contrast to traditional models that solely rely on temperature data, the proposed model utilizes thermal image data as a key input parameter and incorporates temperature data from sensitive points to fully represent the machine's temperature distribution. Furthermore, the attention mechanism is used to optimize the hyperparameters and network structure of the residual network model. Transfer learning is employed to improve training efficiency, reduce data requirements, and enhance the model's transferability. The optimized model achieves a prediction accuracy of 99.5% and converges more quickly. Finally, thermal error compensation experiments are conducted on the platform of the Siemens 840D system with an average effect of more than 70%. The proposed thermal error compensation method is effective and provides a foundation for precision machining. [ABSTRACT FROM AUTHOR]

Despite significant strides in digital classification of Chinese ink paintings, existing methods predominantly rely on low-level features, insufficient for capturing the nuanced artistic styles of such works. This study introduces a novel multi-branch residual network that leverages depth correlation features to enhance the classification of Chinese ink paintings. We innovatively combine global style features, extracted using the Gram matrix, with local brushstroke features obtained via the Holistically-Nested Edge Detection(HED) method. This dual-feature approach addresses the limitations of previous studies by incorporating high-level stylistic nuances alongside low-level details, resulting in a more robust classification system. Quantitative results demonstrate a marked improvement in classification accuracy, with our network outperforming existing state-of-the-art models by significant margins in both artist and genre classification tasks. This advancement not only underscores the efficacy of integrating diverse feature sets but also paves the way for more sensitive and accurate management of digital art repositories. [ABSTRACT FROM AUTHOR]

Copyright of Machine Tool & Hydraulics is the property of Guangzhou Mechanical Engineering Research Institute (GMERI) and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Autism spectrum disorder (ASD) is a disorder in neurological growth, which includes cognitive and behavioral impairment and it starts from infancy. However, the reason for ASD is still vague and no effective medical ways are used for its discovery. Earlier discovery of ASD is extensively beneficial for children's health sustainability. The classical detection models rely on expertise analysis which tends to be expensive and inaccurate. Thus, this paper presents an effectual autism diagnostic model with electroencephalogram (EEG) signals that are produced through the electrical activities of the brain for detecting ASD. The Gaussian filter is employed to abandon the noise. Various statistical features signal and spectral-based features are extracted and provided to DRN for enhanced efficiency. The ASD detection is undergone using Chronological Sewing Training Optimization-Deep Residual Network (CSTO-DRN) wherein DRN is pre-trained using CSTO algorithm by tuning finest weights. The CSTO is built by incorporating the Chronological concept with Sewing Training-Based Optimization (STBO). The CSTO-DRN provided finest accuracy of 88.6%, Negative Predictive Value (NPV) of 87.8%, Positive Predictive Value (PPV) of 89.4%, True negative rate (TNR) of 85%, True positive rate (TPR) of 88.9%, and F-Measure of 87.5%. Its execution can enhance the efficiency of detection and minimize cost and human intervention. [ABSTRACT FROM AUTHOR]

Power distribution networks with distributed generation (DG) face challenges in fault diagnosis due to the high uncertainty, randomness, and complexity introduced by DG integration. This study proposes a two-stage approach for fault location and identification in distribution networks with DG. First, an improved bald eagle search algorithm combined with the Dijkstra algorithm (D-IBES) is developed for fault location. Second, a fusion deep residual shrinkage network (FDRSN) is integrated with IBES and support vector machine (SVM) to form the FDRSN-IBS-SVM model for fault identification. Experimental results showed that the D-IBES algorithm achieved a CPU loss rate of 0.54% and an average time consumption of 1.70 seconds in complex scenarios, outperforming the original IBES algorithm. The FDRSN-IBS-SVM model attained high fault identification accuracy (99.05% and 98.54%) under different DG output power levels and maintained robustness (97.89% accuracy and 97.54% recall) under 5% Gaussian white noise. The proposed approach demonstrates superior performance compared to existing methods and provides a promising solution for intelligent fault diagnosis in modern distribution networks. [ABSTRACT FROM AUTHOR]

Recently, learning through multi-classifiers is of huge interest in economic as well as industrial domains. In addition, the neural network becomes an emerging technique for learning. Nevertheless, the accuracy of the neural network is imperfect due to its loss function. Hence, a new cross entropy-based function is devised. The aim is to develop a model to classify IoT services and build a semantic graph network using a fractional cross entropy-based loss function (FCEBLF). Originally, the service recipe of IFTTT (if-this-then-that) is considered to extract the title and description. Here, the edge and nodes are used for constructing a semantic graph, where semantic features are obtained. Meanwhile, term frequency and inverse document frequency (TF-IDF) are accomplished using IFTTT recipes. Moreover, the natural language processing (NLP) features are extracted to further increase the efficiency. In addition, the obtained features are fused. Thus, the fusion of features is executed by using deep neural network (DNN) based on Hellinger distance. Lastly, the classification of IoT service is attained based on deep residual network (DRN) in which loss function is enhanced using FCEBLF. The proposed FCEBLF + DRN outperformed with high micro-F1 of 91.2%, precision of 91%, and recall of 91.4%. [ABSTRACT FROM AUTHOR]

Vehicular Ad hoc Network (VANET) has recently gained significant attention as a means of enhancing the mobility, efficiency, and safety of applications in the intelligent transportation system. However, because of its high-speed mobility, wireless connectivity, and extensive node coverage, security is a more difficult procedure. The Sybil security threat on VANET is a growing problem today. The Road Side Unit (RSU) failed to synchronise its clock with the legal vehicle, then unplanned vehicles are predicted, thereby incorrect messages are transferred to them. In this paper, Competitive Dolphin Echolocation Optimisation (CDEO)-based Deep Residual Network is proposed for Sybil attack and RSU misbehaviour detection. Here, the effective routing process is performed using Fractional Glow-Worm Swarm Optimisation (FGWSO)-based traffic-aware routing protocol. In the base station, the Sybil attack detection is done. The Sybil attack detection process is done using a Deep residual network, which is trained by the proposed CDEO algorithm. The CDEO algorithm is devised by incorporating Dolphin Echolocation Optimisation (DEO) technique and Competitive Swarm Optimiser (CSO). Additionally, using the Deep residual network, the RSU misbehaviour detection is done. The performance of the developed method is compared with certain performance metrics, like precision, F1-measure, and recall of 0.9197, 0.9121, and 0.9046. [ABSTRACT FROM AUTHOR]

The pandemic disease Coronavirus 2019 (COVID-19) caused thousands of infections and deaths globally. It is important to introduce new medicines to address the critical situation in the medical system. The determination of approximate pIC value is necessary for designing medicines based on molecular compounds. Generally, the approximation of pIC value is a lengthy process, so it is difficult and time-consuming. Hence it is essential to introduce a new technique for automatic approximation. In this research, a Convolutional Neural Network-based transfer learning (CNN-TL) is designed for approximating the pIC value. Initially, Simplified Molecular Input Line Entry System (SMILES) notation is extracted from SMILES string symbols using an entropy-based one-hot encoding matrix and the molecular formula-based encoding. The molecular features are then extracted from the input data using Lorentzian similarity and Deep Residual Network (DRN). The pIC value approximation is performed using the CNN-TL model, where the Visual Geometry Group Network-16 (VGGNet-16) is used to fetch hyperparameters used to initialize the CNN. The experimental results proved that the designed CNN-TL technique achieved minimum error rates with normalized values of 0.406 for R2, 0.516 for Root Mean Square Error (RMSE), 0.267 for Mean Square Error (MSE), and for 0.277 Mean Absolute Percentage Error (MAPE). [ABSTRACT FROM AUTHOR]

Wearable devices are paramount in health monitoring applications since they provide contextual information to identify and recognize human activities. Although sensor-based human activity recognition (HAR) has been thoroughly examined, prior studies have yet to definitively differentiate between symmetric and asymmetric motions. Determining these movement patterns might provide a more profound understanding of assessing physical activity. The main objective of this research is to investigate the use of wearable motion sensors and deep convolutional neural networks in the analysis of symmetric and asymmetric activities. This study provides a new approach for classifying symmetric and asymmetric motions using a deep residual network incorporating channel and spatial convolutional block attention modules (CBAMs). Two publicly accessible benchmark HAR datasets, which consist of inertial measurements obtained from wrist-worn sensors, are used to assess the model's efficacy. The model we have presented is subjected to thorough examination and demonstrates exceptional accuracy on both datasets. The ablation experiment examination also demonstrates noteworthy contributions from the residual mappings and CBAMs. The significance of recognizing basic movement symmetries in increasing sensor-based activity identification utilizing wearable devices is shown by the enhanced accuracy and F1-score, especially in asymmetric activities. The technique under consideration can provide activity monitoring with enhanced accuracy and detail, offering prospective advantages in diverse domains like customized healthcare, fitness tracking, and rehabilitation progress evaluation. [ABSTRACT FROM AUTHOR]

The intelligent monitoring of cutting tools used in the manufacturing industry is steadily becoming more convenient. To accurately predict the state of tools and tool breakages, this study proposes a tool wear prediction technique based on multi-sensor information fusion. First, the vibrational, current, and cutting force signals transmitted during the machining process were collected, and the features were extracted. Next, the Kalman filtering algorithm was used for feature fusion, and a predictive model for tool wear was constructed by combining the ResNet and long short-term memory (LSTM) models (called ResNet-LSTM). Experimental data for thin-walled parts obtained under various machining conditions were utilized to monitor the changes in tool conditions. A comparison between the ResNet and LSTM tool wear prediction models indicated that the proposed ResNet-LSTM model significantly improved the prediction accuracy compared to the individual LSTM and ResNet models. Moreover, ResNet-LSTM exhibited adaptive noise reduction capabilities at the front end of the network for signal feature extraction, thereby enhancing the signal feature extraction capability. The ResNet-LSTM model yielded an average prediction error of 0.0085 mm and a tool wear prediction accuracy of 98.25%. These results validate the feasibility of the tool wear prediction method proposed in this study. [ABSTRACT FROM AUTHOR]

To improve the accuracy and efficiency of tool wear predictions, this study proposes a tool wear prediction model called LSTM_ResNet which is based on the long short-term memory (LSTM) network and the Residual Network (ResNet). The model utilizes LSTM layers for processing, where the first block and loop blocks serve as the core modules of the deep residual network. The model employs a series of methods including convolution, batch normalization (BN), and Rectified Linear Unit (ReLU) to enhance the model's expression and prediction capabilities. The performance of the LSTM_ResNet model was evaluated using experimental data from the PHM2010 datasets and two different depths (64 and 128 layers), training both LSTM_ResNet models for 200 epochs. The 64-layer model's root mean square error (RMSE) values are 3.36, 4.35, and 3.59, and the mean absolute error (MAE) values are 2.42, 2.85, and 2.21; using 128 layers, the RMSE values are 3.66, 3.99, and 3.77, and the MAE values are 2.49, 2.73, and 3.01. The results indicate that the 64-layer LSTM has smaller average errors, suggesting that compared to other common network structures, the LSTM_ResNet network has a higher performance. This research provides an effective solution for tool wear prediction and helps to improve the technical level of tool wear prediction in China. [ABSTRACT FROM AUTHOR]

The planning, scheduling and operation decisions of the power grid depend on the calculation of the simulation model. Parameter errors in the grid model can lead to deviations between simulation calculations and actual grid operation. The strategy based on incorrect calculation data will lead to power outages in the actual power grid, which may cause significant economic losses and personal safety accidents. For the safe operation of power grid, a method for locating the wrong parameters of transmission line based on sensitivity analysis (SA) and deep residual network (DRN) is proposed. By calculating the sensitivity of apparent power to parameter error of each line in the power grid, the propagation characteristics of power flow error are analyzed quantitatively. An error region segmentation method is proposed to reduce the search range of error parameters from large-scale power grids to local networks which can reduce the computational complexity of the search algorithm, and increase accuracy. An error transmission line index for local power grids is proposed to identify error source in local power grids. Then, the specific wrong parameters are identified through the DRN. It can intelligently identify the error parameters from multiple parameters of the error transmission line. The calculation results of the 300-bus system verify the correctness and effectiveness of the proposed method. [ABSTRACT FROM AUTHOR]

During the construction process of tunnels, the cutterhead of shield tunneling machines may get clogged due to clay adhesion, which may seriously affect the efficiency of the project. Therefore, finding an intelligent diagnosis method to detect the clogging status is of great importance. In this study, a deep residual network-based method for diagnosing cutterhead clogging on shield tunneling machines is proposed. First, working state data of the shield tunneling machine is screened out, and parameters reflecting the clogging state are selected for further analysis. After eliminating extreme outliers, an empirical formula is proposed to label the data. At the same time, several time-domain features of the selected excavation parameters within every five minutes are extracted. These features are then fed into the proposed model as the input data to realize clogging detection. Because the original dataset is unbalanced, the combination of f1-score and accuracy is used to evaluate the performance of the proposed model. The results show that the accuracy of the proposed algorithm reaches 95.71%, which is 1.21%, 2.84%, 9.84%, 6.04%, and 0.86% higher than the support vector machine-based, random forest-based, AdaBoost-based, extreme gradient boosting-based and deep neural network-based methods. The f1 score of the proposed model is 0.923, which is also 0.038, 0.042, 0.269, 0.169 and 0.02 higher than those compared methods. Therefore, the proposed deep residual network-based method can accurately detect cutterhead clogging conditions. [ABSTRACT FROM AUTHOR]

Accurate and appropriate prostate cancer detection can significantly reduce the death rate. In this research, the Henry Firefly Gas Solubility Optimization (HFGSO)-based Deep Residual Network (DRN) is established for the autonomous detection of prostate cancer. The pre-processing is done by Cuckoo Search-Based (T2FCS) filter and Type 2 Fuzzy. Subsequently, segmentation is exhibited by devised multi-objective SegNet scheme. The multi-objective SegNet method is newly designed by updating the objective function of SegNet with loss function. The multi-objective SegNet is trained by HFGSO. Then, data augmentation is done with cropping and rotation, which improves the performance of detection. At last, cancer identification is executed with DRN, and it is trained by HFGSO. The developed optimized multi-objective SegNet with DRN technique also achieved increased performance for the detection of cancer, with a sensitivity, specificity, and accuracy 0.9367, 0.9130, and 0.9263. [ABSTRACT FROM AUTHOR]

Fingermarks play an important role in document identification. At the same time, fingermarks on paper documents are often accompanied by signatures and background text, which introduce noise to the original fingermark textures and increase the difficulty of detection. A signed fingermark detection method based on deep residual networks and a decision-level fusion strategy was proposed to defend against spoofing attacks from fake fingermarks. Firstly, the multi-scale structure was introduced in the residual module, which improved the network's depth and breadth without increasing the parameters. Then, the multi-probability label strategy was refined and employed to enhance the local encoding ability of the feature extraction. A score fusion strategy was designed, with weights allocated based on the difference in signed interference levels of local image blocks. Finally, a model fusion strategy based on evidence theory was suggested, which improved detection accuracy by leveraging complementarity between models. A large-scale fingermark database was established, which included real fingermarks made from real fingers and fake fingermarks made from various materials, and this was divided into two sub databases: signed and unsigned. The experimental results show that the proposed method achieves 96.16% accuracy based on the fingerprint dataset of the global liveness detection competition called LivDet2017 and achieves 99.30% accuracy based on the signed fingermark database, while it has good resistance to spoofing attacks from unknown materials. [ABSTRACT FROM AUTHOR]

The development of deep residual network (ResNet) has contributed significantly to the progress of computer vision and image classification, expanding the applicability of convolutional neural networks to different fields. Researchers continue to improve the classification accuracy of ResNet by increasing parameter sizes or model complexity. However, enlarging the network parameter size would significantly increase the training workload. In this work, we adjusted the scale and variation pattern of ResNet18 channel numbers, evaluated their performance differences using different datasets, and designed a flask-like channel structure, which enabled ResNet18 to reduce model parameters while maintaining accuracy. Then, we use MNIST and STL10 datasets validate the effectiveness of FLC structure. Finally, we extend the FLC structure to other ResNet models with different layers, such as ResNet34, ResNet50, ResNet101, and ResNeXt. By testing these ResNet models on the CIFAR10 dataset, our experiments showed that the ResNet models with the FLC structure (namely ResNet_FLC) can maintain or improve the accuracy of the model by approximately 1% while reducing the number of model parameters and FLOPs.

Epileptic seizures are a prevalent neurological condition that impacts a considerable portion of the global population. Timely and precise identification can result in as many as 70% of individuals achieving freedom from seizures. To achieve this, there is a pressing need for smart, automated systems to assist medical professionals in identifying neurological disorders correctly. Previous efforts have utilized raw electroencephalography (EEG) data and machine learning techniques to classify behaviors in patients with epilepsy. However, these studies required expertise in clinical domains like radiology and clinical procedures for feature extraction. Traditional machine learning for classification relied on manual feature engineering, limiting performance. Deep learning excels at automated feature learning directly from raw data sans human effort. For example, deep neural networks now show promise in analyzing raw EEG data to detect seizures, eliminating intensive clinical or engineering needs. Though still emerging, initial studies demonstrate practical applications across medical domains. In this work, we introduce a novel deep residual model called ResNet-BiGRU-ECA, analyzing brain activity through EEG data to accurately identify epileptic seizures. To evaluate our proposed deep learning model’s efficacy, we used a publicly available benchmark dataset on epilepsy. The results of our experiments demonstrated that our suggested model surpassed both the basic model and cutting-edge deep learning models, achieving an outstanding accuracy rate of 0.998 and the top F1-score of 0.998.

Human activity recognition (HAR) in real-world settings has gained significance due to the growth of Internet of Things (IoT) devices such as smartphones and smartwatches. Nonetheless, limitations such as fluctuating environmental conditions and intricate behavioral patterns have impacted the accuracy of the current procedures. This research introduces an innovative methodology employing a modified deep residual network, called 1D-ResNeXt, for IoT-enabled HAR in uncontrolled environments. We developed a comprehensive network that utilizes feature fusion and a multi-kernel block approach. The residual connections and the split–transform–merge technique mitigate the accuracy degradation and reduce the parameter number. We assessed our suggested model on three available datasets, mHealth, MotionSense, and Wild-SHARD, utilizing accuracy metrics, cross-entropy loss, and F1 score. The findings indicated substantial enhancements in proficiency in recognition, attaining 99.97% on mHealth, 98.77% on MotionSense, and 97.59% on Wild-SHARD, surpassing contemporary methodologies. Significantly, our model attained these outcomes with considerably fewer parameters (24,130–26,118) than other models, several of which exceeded 700,000 parameters. The 1D-ResNeXt model demonstrated outstanding effectiveness under various ambient circumstances, tackling a significant obstacle in practical HAR applications. The findings indicate that our modified deep residual network presents a viable approach for improving the dependability and usability of IoT-based HAR systems in dynamic, uncontrolled situations while preserving the computational effectiveness essential for IoT devices. The results significantly impact multiple sectors, including healthcare surveillance, intelligent residences, and customized assistive devices.

This study aims to capture subtle changes in the pupil, identify relatively weak inter-class changes, extract more abstract and discriminative pupil features, and study a pupil refinement recognition method based on attention mechanisms. Based on the deep learning framework and the ResNet101 deep residual network as the backbone network, a pupil refinement recognition model is established. Among them, the image preprocessing module is used to preprocess the pupil images captured by infrared spectroscopy, removing internal noise from the pupil images. By using the ResNet101 backbone network, subtle changes in the pupil are captured, weak inter-class changes are identified, and different features of the pupil image are extracted. The channel attention module is used to screen pupil features and obtain key pupil features. External attention modules are used to enhance the expression of key pupil feature information and extract more abstract and discriminative pupil features. The Softmax classifier is used to process the pupil features captured by infrared spectra and output refined pupil recognition results. Experimental results show that this method can effectively preprocess pupil images captured by infrared spectroscopy and extract pupil features. This method can effectively achieve fine pupil recognition, and the fine recognition effect is relatively good.

Non-technical losses (NTL) refer to unrecorded power consumption generated by dishonest customers, which is a substantial issue affecting the power system stability and economic efficiency of the power grid. The detection of dishonest customers is hindered by the complexity of NTL such as data feature selection, retention time feature, and power consumption pattern judgment. This work addresses these issues using meter recording data, proposing an NTL detection approach with Gram’s angle field (GAF) and deep residual network (ResNet). Principal component analysis (PCA) method is applied to compress multiple electricity detection indexes, which aims to obtain multi-dimensional power consumption data characteristics without changing its timing characteristics. The GAF method is used to convert the time-series power features of individual users into a two-dimensional image, achieving the purpose of maintaining the user’s time-series features and user-based units. The images generated by the GAF method, which contain information about the electricity consumption characteristics of many customers, are classified by ResNet to highlight customers with NTL. The claimed algorithm was tested on a dataset consisting of both fraudulent and non-fraudulent subscriber data. The results demonstrated that the NTL detection method based on GAF and ResNet is superior to the traditional NTL detection method and has high accuracy.

The collaborative robot can complete various drilling tasks in complex processing environments thanks to the high flexibility, small size and high load ratio. However, the inherent weaknesses of low rigidity and variable rigidity in robots bring detrimental effects to surface quality and drilling efficiency. Effective online monitoring of the drilling quality is critical to achieve high performance robotic drilling. To this end, an end-to-end drilling-state monitoring framework is developed in this paper, where the drilling quality can be monitored through online-measured vibration signals. To evaluate the drilling effect, a Canny operator-based edge detection method is used to quantify the inclination state of robotic drilling, which provides the data labeling information. Then, a robotic drilling inclination state monitoring model is constructed based on the Resnet network to classify the drilling inclination states. With the aid of the training dataset labeled by different inclination states and the end-to-end training process, the relationship between the inclination states and vibration signals can be established. Finally, the proposed method is verified by collaborative robotic drilling experiments with different workpiece materials. The results show that the proposed method can effectively recognize the drilling inclination state with high accuracy for different workpiece materials, which demonstrates the effectiveness and applicability of this method. [ABSTRACT FROM AUTHOR]

In the context of mountain tunnel mining through the drilling and blasting method, the recognition of lithology from palm face images is crucial for the comprehensive analysis of geological conditions and the prevention of geological risks. However, the complexity of the background in the acquired palm face images, coupled with an insufficient data sample size, poses challenges. While the incorporation of deep learning technology has enhanced lithology recognition accuracy, issues persist, including inadequate feature extraction and suboptimal recognition accuracy. To address these challenges, this paper proposes a lithology recognition network integrating attention mechanisms and a feature Brownian distance covariance approach. Drawing inspiration from the Brownian distance covariance concept, a feature Brownian distance covariance module is devised to enhance the network's attention to rock sample features and improve classification accuracy. Furthermore, an enhanced lightweight Convolutional Block Attention Module is introduced, with upgrades to the multilayer perceptron in the channel attention module. These improvements emphasize attention to lithological features while mitigating interference from background information. The proposed method is evaluated on a palm face image dataset collected in the field. The proposed method was evaluated on a dataset comprising field-collected images of a tunnel rock face. The results illustrate a significant enhancement in the improved model's ability to recognize rock images, as evidenced by improvements across all objective evaluation metrics. The achieved accuracy rate of 97.60% surpasses that of the current mainstream lithology recognition neural network. [ABSTRACT FROM AUTHOR]

The completion degree of sports training can not reach the corresponding standard, and the training effect will be greatly weakened. In order to improve the effect of sports training, the evaluation method of sports training completion degree based on deep residual network is studied. The image collector based on ARM is used to collect the action images of athletes in sports training, and the collected action images are preprocessed based on spatial scale filtering and regression factors. Construct a depth residual network, learn the implicit relationship between athletes' state and the dynamic change process of sports training actions through off-line training, and train the model; In the online application process, the preprocessed action images will be input into the trained evaluation model to evaluate the athletes' sports training action completion in real time. At the same time, residual shrinkage unit and attention mechanism are used to optimize the depth residual network, which improves the training efficiency and evaluation performance of the network. The experimental results show that this method has good evaluation performance under the condition of setting parameters, and can effectively improve the effect of physical training. [ABSTRACT FROM AUTHOR]

Encrypted traffic identification pertains to the precise acquisition and categorization of data from traffic datasets containing imbalanced and obscured content. The extraction of encrypted traffic attributes and their subsequent identification presents a formidable challenge. The existing models have predominantly relied on direct extraction of encrypted traffic data from imbalanced datasets, with the dataset's imbalance significantly affecting the model's performance. In the present study, a new model, referred to as UD-VLD (Unbalanced Dataset-VAE-LSTM-DRN), was proposed to address above problem. The proposed model is an encrypted traffic identification model for handling unbalanced datasets. The encoder of the variational autoencoder (VAE) is combinedwith the decoder and Long-short term Memory (LSTM) in UD-VLD model to realize the data enhancement processing of the original unbalanced datasets. The enhanced data is processed by transforming the deep residual network (DRN) to address neural network gradient-related issues. Subsequently, the data is classified and recognized. The UD-VLD model integrates the related techniques of deep learning into the encrypted traffic recognition technique, thereby solving the processing problem for unbalanced datasets. The UD-VLD model was tested using the publicly available Tor dataset and VPN dataset. The UD-VLD model is evaluated against other comparativemodels in terms of accuracy, loss rate, precision, recall, F1-score, total time, and ROC curve. The results reveal that the UD-VLD model exhibits better performance in both binary and multi classification, being higher than other encrypted traffic recognition models that exist for unbalanced datasets. Furthermore, the evaluation performance indicates that the UD-VLD model effectivelymitigates the impact of unbalanced data on traffic classification. and can serve as a novel solution for encrypted traffic identification. [ABSTRACT FROM AUTHOR]

As the degradation factors of remote sensing images become increasingly complex, it becomes challenging to infer the high-frequency details of remote sensing images compared to ordinary digital photographs. For super-resolution (SR) tasks, existing deep learning-based single remote sensing image SR methods tend to rely on texture information, leading to various limitations. To fill this gap, we propose a remote sensing image SR algorithm based on a multi-scale texture transfer network (MTTN). The proposed MTTN enhances the texture feature information of reconstructed images by adaptively transferring texture information according to the texture similarity of the reference image. The proposed method adopts a multi-scale texture-matching strategy, which promotes the transmission of multi-scale texture information of remote sensing images and obtains finer-texture information from more relevant semantic modules. Experimental results show that the proposed method outperforms state-of-the-art SR techniques on the Kaggle open-source remote sensing dataset from both quantitative and qualitative perspectives. [ABSTRACT FROM AUTHOR]

The accurate prediction of the rate of penetration (ROP) is crucial for optimizing drilling parameters and enhancing drilling efficiency in ultra-deep wells. However, this task is challenging due to the harsh geological conditions, complex drilling processes, voluminous drilling data, and nonlinear relationships between drilling parameters and rock-breaking. In this study, a comprehensive intelligent model is proposed that combines clustering and deep residual neural network to address these challenges. Specifically, relevant feature parameters are selected for ROP prediction and the Savitzky–Golay filter is employed to reduce noise in the field data. Formations with similar rock characteristics are clustered using well logging parameters, including sonic logging and natural gamma ray logging, which indicate the formation rock properties. A deep residual neural network is then used to develop the prediction model, with the clustering results and 13 mud logging parameters serving as inputs. The model is trained and tested using field data from an ultra-deep reservoir in northwest China, and its performance is evaluated. The impact of data noise reduction, formation clustering, and deep residual neural network on the prediction accuracy is analyzed through ablation experiments. The proposed model achieves high accuracy in predicting ROP, with relative errors ranging from 11.34 to 11.44% and R2 values from 0.92 to 0.94. Compared to traditional machine learning models, the approach demonstrates superior performance and is suitable for real-time drilling applications. This study provides a promising solution for accurate ROP prediction in ultra-deep wells, helping to optimize drilling parameters and improve drilling efficiency. [ABSTRACT FROM AUTHOR]

Epileptic seizures are a prevalent neurological condition that impacts a considerable portion of the global population. Timely and precise identification can result in as many as 70% of individuals achieving freedom from seizures. To achieve this, there is a pressing need for smart, automated systems to assist medical professionals in identifying neurological disorders correctly. Previous efforts have utilized raw electroencephalography (EEG) data and machine learning techniques to classify behaviors in patients with epilepsy. However, these studies required expertise in clinical domains like radiology and clinical procedures for feature extraction. Traditional machine learning for classification relied on manual feature engineering, limiting performance. Deep learning excels at automated feature learning directly from raw data sans human effort. For example, deep neural networks now show promise in analyzing raw EEG data to detect seizures, eliminating intensive clinical or engineering needs. Though still emerging, initial studies demonstrate practical applications across medical domains. In this work, we introduce a novel deep residual model called ResNet-BiGRU-ECA, analyzing brain activity through EEG data to accurately identify epileptic seizures. To evaluate our proposed deep learning model's efficacy, we used a publicly available benchmark dataset on epilepsy. The results of our experiments demonstrated that our suggested model surpassed both the basic model and cutting-edge deep learning models, achieving an outstanding accuracy rate of 0.998 and the top F1-score of 0.998. [ABSTRACT FROM AUTHOR]