High Recognition Rate Research Articles

Self-rectifying NiOX/WOX heterojunction synaptic memristor for crossbar architectured reservoir computing system

In this study, we examine an ITO/NiOX/WOX/Pt p-n heterojunction memristor for neuromorphic applications as a synaptic crossbar array. The transition in the depletion region at the interface between p-type NiOX and n-type WOX crucially influences the self-rectifying characteristics of the device depending on the voltage polarity. Furthermore, long- and short-term memory coexist depending on the switching voltage condition, thus enabling various neuromorphic applications, such as reservoir computing and the use of the synaptic device at the off-chip trained network. The reliable operational characteristics are confirmed by obtaining an average memory window (>9) and rectification ratio (>58) between device-to-device and cycle-to-cycle. Furthermore, synaptic functions were successfully implemented, such as spike-rate-dependent plasticity, spike-number-dependent plasticity, paired-pulse facilitation, post-tetanic potentiation, paired-pulse depression, and post-tetanic depression, in conjunction with repeatability. Ultimately, reservoir computing is accomplished based on the I–V nonlinearity and short-term memory characteristics. A high pattern recognition rate (>96.2 %) is achieved in MNIST tasks using 16 reservoir states, thus affirming the trustworthiness of the reservoir computing system. With its comprehensive approach to synaptic applications for neuromorphic computing systems, the heterojunction device highlights its considerable potential to advance artificial neural networks in conjunction with novel memristor technology directions.

Recognition of marine navigators’ workload based on eye movement features using bridge simulation

The human factor accounts for 80% of marine traffic accidents, either directly or indirectly. Workload significantly impacts the reliability of marine navigators’ behavior. Employing eye movement measurement as an objective assessment tool can effectively evaluate seafarers’ workload. This study conducts bridge simulation experiments to analyze eye movement features and investigate marine navigators’ workload levels across different tasks. By categorizing marine navigators’ workload into high and low level based on specific navigation duties. The study employs an independent sample T-test to select blink frequency, fixation time percentage, and pupil area standard deviation as key eye movement features for workload recognition. A marine navigators’ workload recognition model is developed utilizing the support vector machine (SVM) algorithm, which is subsequently trained and evaluated with experimental data. Results demonstrate a high recognition rate of 90.91% for marine navigators’ workload, indicating the SVM method’s accuracy in workload assessment based on eye movement features. These findings lay the foundation for the advancement of intelligent monitoring systems for marine navigators’ workload and the implementation of human–computer interaction active early warning technology and equipment.

Hyperspectral imaging coupled with deep learning model for visualization and detection of early bruises on apples

Early bruise on apples caused by external impacts during the transportation process is commonly difficult to be detected on the apple surface, limiting the application of traditional machine vision methods in determining fruit quality. In recent years, hyperspectral imaging (HSI) has emerged as a promising technology for identifying early bruise of fruits due to its efficient and nondestructive detection. In this study, HSI data in the shortwave infrared range were collected at 2-hour and 6-hour intervals after mechanical damage. The combination of the successive projections algorithm (SPA) and principal component analysis (PCA) was used to select three key feature bands, namely, 1074 nm, 1269 nm and 1441 nm. Pseudo color transformation and band ratio algorithm were then employed to improve the contrast between damaged and healthy apple tissues for image enhancement. The fast and precise YOLOv5 (FP-YOLOv5) model achieved effective identification of apple bruises, with a high recognition rate of 95 % and a fast detection speed at 130 fps. Overall, the proposed framework based on band selection and image enhancement exhibits better performance in the detection of early apple bruises, providing useful insights for HSI combined with a deep learning model in the grading evaluation of fruit quality.

Detection of Track Bed Defects Based on Fibre Optic Sensor Signals and an Improved Hidden Markov Model

Railway track bed defects affect the normal operation of trains and pose great safety risks. In order to detect such issues early, we developed a railway track bed defect detection method which uses optical fibre sensors and an improved HMM (hidden Markov model) to detect the signals collected by a DAS (distributed acoustic sensing) system. First, by analysing the physical process of train operation and determining the number of hidden states, a waveform segmentation method based on average amplitude was used to solve the problem of unequal signal lengths. Second, an adaptive power spectrum energy ratio calculation method was employed to extract track fault features, a set of which was constructed by combining various quantity features. Then, normal and abnormal models were trained according to the sensor measurement area. Finally, the probability of detecting the signal with each model was compared to determine whether the signal was abnormal. Experiments were conducted to compare the applicability of the waveform segmentation method and the feature extraction method. The results show that the HMM based on both waveform segmentation and track bed defect feature sets had the highest recognition rate, the lowest number of false detection areas, and a greater impact on the signal in the early development stage of track bed defects. The proposed method, therefore, has strong recognition ability, which makes it suitable for track bed defect detection.

Stability and safety study of pumped storage units based on time-shifted multi-scale cosine similarity entropy

The diagnosis of vibration signals of pumped storage units is crucial to the safe and stable operation of the units. In this paper, a fault diagnosis method with high recognition rate and strong generalizability is proposed. Firstly, the original vibration signals are decomposed by Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) decomposition, the decomposed IMF1 components are proposed to be decomposed by Variational Mode Decomposition (VMD) for secondary decomposition, and then based on the fractal theory, a tool to measure the signal complexity is proposed - Time-Shifted Multiscale Cosine Similarity Entropy (TSMCSE). Finally, the entropy value of each component is input into the Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) model as the feature value for diagnosis and identification. Through the measured data of unit operation, four states of 20 MW, 100 MW, 280 MW and 340 MW in the upper guide +X and −Y directions are used for validation, and the results show that the diagnostic accuracy of CEEMDAN-VMD-TSMCSE-CNN-LSTM reaches 100 %. At the same time, in order to experience the generalizability of the model, the open data set of rolling bearings from Jiangnan University is introduced, and four conditions in 600/rpm and 800 rpm are selected for re-validation, and the experimental results show that the accuracy of the model designed in this paper is 100 %. The method provides a new tool for the study of stability and safety of pumped storage units.

Research on image recognition of tomato leaf diseases based on improved AlexNet model

Aiming at the problems that the traditional image recognition technology is challenging to extract useful features and the recognition time is extended; the AlexNet model is improved to improve the effect of image classification and recognition. This study focuses on 8 types of tomato leaf diseases and healthy leaves. By using HOG and LBP weighted fusion to extract image features, a tomato leaf disease recognition model based on the AlexNet model is proposed, and transfer learning is used to train the AlexNet model. Transfer the knowledge learned by the AlexNet model on the PlantVillage image dataset to this model while reducing the number of fully connected layers. Keras deep learning framework and programming language Python were used. The model was implemented, and the classification and identification of tomato leaf diseases were carried out. The recognition rate of feature-weighted fusion classification is higher than that of serial and parallel methods, and the recognition time is the shortest. When the weight coefficient ratio of HOG and LBP is 3:7, the image recognition rate is the highest, and its value is 97.2 %. From the model performance curve See, when the number of iterations is more than 150 times, the training set and test accuracy rate both exceed 97 %, the loss rate shows a gradient decline, and the change is relatively stable; compared with the traditional AlexNet model, HOG + LBP + SVM model, and VGG model, improved AlexNet model has the highest recognition rate, and it has high recall value, accuracy, and F1 value; Compared with the latest convolutional neural network disease recognition models, improved AlexNet model recognition accuracy was 98.83 %, and the F1 value was 0.994. It shows that the model has good convergence performance, fast prediction speed, and low loss rate and can effectively identify 8 types of tomato leaf images, which provides a reference for the research on crop disease identification.

Application of CNN based Regognition Technology in Public English Teaching in College Intelligent Classroom

With the advancement of intelligent technology, CNN-based recognition technology has been integrated into public English teaching in universities. This implementation contributes to enhancing teaching quality, nurturing students’ English proficiency, and holds significant educational and practical value. To address the issue of low traditional attendance efficiency in intelligent classrooms for public English teaching, a face recognition model based on CNN recognition technology has been developed. R-CNN is utilized for object detection, along with pyramid pooling and non-maximum suppression to acquire the optimal candidate region for face detection. Furthermore, K-Means clustering is combined to enhance Fast R-CNN, thereby improving detection accuracy. Experimental results demonstrated that among the three networks - Fast R-CNN, Faster R-CNN, and CNN -Faster R-CNN maintained a high recognition rate and exhibited faster convergence speed, showcasing superior overall performance. Specifically, at 500 iterations, the three networks require 23.7 seconds, 26.8 seconds, and 34.2 seconds, respectively. For facial expression recognition, Faster R-CNN achieved the highest recognition rate, indicating its exceptional detection efficiency and potential for aiding teaching management. This study offers novel technical support for public English teaching in intelligent university classrooms, effectively enhancing teaching efficacy and learning experiences. Its practical significance extends to promoting educational reform and improvement.

Factor analysis and GA-BP-ANN prediction of nitrogen diffusion behavior in underground laboratory under ventilation conditions

Nitrogen is widely used in various laboratories as a suppressive gas and a protective gas. Once nitrogen leaks and accumulates in a such confined space, it will bring serious threats to the experimental staff. Especially in underground tunnels or underground laboratories where there is no natural wind, the threat is more intense. In this work, the ventilation design factors and potential leakage factors are identified by taking the leakage and diffusion of a large liquid nitrogen tank in China Jinping Underground Laboratory (CJPL) as an example. Based on computational fluid dynamics (CFD) research, the effects of fresh air inlet position, fresh air velocity, exhaust outlet position, leakage hole position, leakage hole size, and leaked nitrogen mass flow rate on nitrogen diffusion behavior in specific environments are discussed in detail from the perspectives of nitrogen concentration field and nitrogen diffusion characteristics. The influencing factors are parameterized, and the Latin hypercube sampling (LHS) is used to uniformly sample within the specified range of each factor to obtain samples that can represent the whole sample space. The nitrogen concentration is measured by numerical value, and the nitrogen diffusion characteristics are measured by category. The GA-BP-ANN numerical regression and classification regression models for nitrogen concentration prediction and nitrogen diffusion characteristics prediction are established. By using various rating indicators to evaluate the performance of the trained model, it is found that models have high accuracy and recognition rate, indicating that it is effective in predicting and determining the concentration value and diffusion characteristics of nitrogen according to ventilation factors and potential leakage factors. The research results can provide a theoretical reference for the parametric design of the ventilation system.

Recognition of facial expressions of emotions in tactile drawings by blind children, children with low vision and sighted children

In the context of blindness, studies on the recognition of facial expressions of emotions by touch are essential to define the compensatory touch abilities and to create adapted tools on emotions. This study is the first to examine the effect of visual experience in the recognition of tactile drawings of facial expressions of emotions by children with different visual experiences. To this end, we compared the recognition rates of tactile drawings of emotions between blind children, children with low vision and sighted children aged 6–12 years. Results revealed no effect of visual experience on recognition rates. However, an effect of emotions and an interaction effect between emotions and visual experience were found. Indeed, while all children had a low average recognition rate, the drawings of fear, anger and disgust were particularly poorly recognized. Moreover, sighted children were significantly better at recognizing the drawings of surprise and sadness than the blind children who only showed high recognition rates for joy. The results of this study support the importance of developing emotion tools that can be understood by children with different visual experiences.

Design and analysis of face recognition system based on VGGFace-16 with various classifiers

This research presents a face recognition system based on different classifiers that deal with various face positions. The proposed system involves the extraction of features through the VGG-Face-16 deep neural network, which only extracts essential features of input images, leading to an improved recognition step and enhanced algorithm efficiency, while the recognition involves the radial basis function in support vector machine (SVM) classifier and evaluate the performance of the system. Also, the system is designed and implemented later by using other classifiers; they are K-neareste2 neighbour (KNN) classifiers, logistic regression (LR), gradient boosting (XGBoost), decision tree classifier (DT) and Naive Bayes classifier (NB). The proposed algorithm was tested with the four face databases: AT&T, PINs Face, linear friction welding (LFW) and real database. The database was divided into two groups: One contains a percentage of images that are used for training and the second contains a percentage of images (remainder) which was used for testing. The results show that the classification by RBF in SVM has the highest recognition rate in the case of using small, medium and large databases; it was 100% in AT&T and Real database, while its efficiency appears to be lower when using large-size databases whereas it is 96% in PINs database and 60.1% in LFW database.

A geometric algebra-based approach for myoelectric pattern recognition control and faster prosthesis recalibration

ObjectiveAlthough many advancements have been made on myoelectric pattern-recognition, the control of poly-articulated upper-limb prostheses remains insufficiently robust. Electrode-shift, sweat or fatigue degrade the performance of classifiers over time, resulting in unfruitful device usage and frequent re-calibration. To tackle this issue, here we introduce two models − µP6 and µP8 − that combine Geometric Algebra with nearest-neighbor classification. We aim at reducing both the necessary training data and training time and, unlike most current state-of-the-art algorithms, we exploit an alternative geometric representation (and visualization) of the EMG signal as different polygons for different types of gestures, facilitating the explanation of the decision-making process to a layman. Moreover, we explore four abstention strategies to reduce the number of misclassifications. MethodsWe perform an offline analysis on two datasets, alongside two other standard models: nonlinear logistic regression (NLR) and linear discriminant analysis (LDA). ResultsEven with few training data, the proposed algorithms achieve high F1-scores (>0.95), significantly higher or non-significantly different from the values obtained with NLR and LDA, while maintaining relatively low abstentions rates and training times (<2 ms). ConclusionThe proposed algorithms allow to reduce the amount of training data and training times without compromising recognition rates. SignificanceThe proposed algorithms may contribute for a faster prosthesis re-calibration procedure while allowing to re-gain high recognition rates. Furthermore, the decision-making process is explainable and interpretable, potentially improving user trust and acceptance.

Neuromorphic Computing in Synthetic Antiferromagnets by Spin‐Orbit Torque Induced Magnetic‐Field‐Free Magnetization Switching

AbstractSynthetic antiferromagnet (SAF) with high thermal stability, ultra‐fast spin dynamics, and highly efficient spin‐orbit torque switching has great application potential in neuromorphic computing hardware. However, two challenges, the weakening of Hall signal in the remanent state and the need for a large auxiliary magnetic field for perpendicular magnetization switching, greatly limit the advantages of SAF in neuromorphic computing. In this work, both the enhanced anomalous Hall resistance and magnetic‐field‐free perpendicular magnetization switching are achieved by using oblique sputtering to fabricate the Pt/CoPt/Ru/CoTb SAF with strong interlayer exchange coupling and magnetic moment compensation. The fabricated SAF as synapse shows nearly linear, nonvolatile multistate plasticity, and as neuron exhibits a nonlinear sigmoid activation function, which are used to construct a fully connected neural network with a remarkable 97.0–98.1% recognition rate for the handwritten digits. Additionally, SAF serving as spike‐timing‐dependent plasticity synapse is used to construct an adaptive, unsupervised learning spiking neural network, and achieve an 87.0% accuracy in handwritten digit recognition. The findings exhibit the promise of SAFs as specialized hardware for high‐performance neuromorphic computing, offering high recognition rates and low power consumption.

High sensitivity artificial synapses using printed high-transmittance ITO fibers for neuromorphic computing

Artificial synapses are essential building blocks for neuromorphic electronics. Here, solid polymer electrolyte-gated artificial synapses (EGASs) were fabricated using ITO fibers as channels, which possess an ultra-high sensitivity of 5 mV and a long-term memory time exceeding 3 min. Notably, digitally printed ITO-fiber arrays exhibit an ultra-high transmittance of approximately 99.67 %. Biological synaptic plasticity, such as excitatory postsynaptic current, paired-pulse facilitation, spike frequency-dependent plasticity, and synaptic potentiation and depression, were successfully mimicked using the EGASs. Based on the synaptic properties of the EGASs, an artificial neural network was constructed to perform supervised learning using the Fashion-MNIST dataset, achieving high pattern recognition rate (82.39 %) due to the linear and symmetric synaptic plasticity. This work provides insights into high-sensitivity artificial synapses for future neuromorphic computing.

Research on rotor fault diagnosis technology of three-phase asynchronous motor based on NA-MEMD mutual information and SVM

Aiming at the problem of mode aliasing in the adaptive decomposition of nonlinear and non-stationary current signals generated by three-phase asynchronous motor faults, and the fault features contained in signals collected by a single sensor can not be accurately and comprehensively extracted and characterized when early rotor bar breakage and air gap eccentricity faults occur, A fault diagnosis method for three-phase asynchronous motor based on noise assisted multivariate empirical mode decomposition (NA-MEMD) and mutual information is proposed. Firstly, the NA-MEMD algorithm is used to decompose the three-phase stator current signal of the asynchronous motor to obtain multi-scale intrinsic mode functions (IMFs). Then, the correlation algorithm is used to screen the IMFs containing useful information. Then, the filtered IMF components are reconstructed into new signals and their features are extracted, Finally, support vector machines (SVM) are used to identify the rotor broken bars and air gap eccentric faults of the three-phase asynchronous motor. The experimental results show that the NA-MEMD method has a higher recognition rate than the traditional empirical mode decomposition (EMD) and the ensemble empirical mode decomposition (EEMD) methods.

Research on fault location of DC line in UHV converter station based on digital twin technology

The technologies such as data, information and intelligence are gradually applied to the power industry, and the power industry is gradually realizing digital transformation to meet the requirements of more reliable, stable, intelligent and safe power supply. UHV converter station is the heart of power grid equipment, which mainly ensures the conversion between AC and DC to ensure the stable operation of the power system. Facing the situation that the fault location is difficult and the loss is too large, this paper introduces the Digital Twin Technology(DTT), which completes the real-time mapping of the data state between the physical body and the virtual body of the converter station line through the digital modeling of the physical intelligent equipment. The recognition rate of CNN-BiGRU model for line fault and type reaches 99.78 % and 98.56 %. The CNN-BiGRU model has the highest recognition rate and the best effect in terms of the line recognition rate and the fault type recognition rate. The maximum Recall and Precision values of CNN-BiGRU model are 99.67 % and 97.96 %. The closer the F1 is to 1, the better the performance of the model is, and the value of CNN-BiGRU reaches 0. 94, which shows that the model has good performance.

Rotating machinery fault classification based on one-dimensional residual network with attention mechanism and bidirectional gated recurrent unit

Abstract Conventional convolutional neural networks (CNNs) predominantly emphasize spatial features of signals and often fall short in prioritizing sequential features. As the number of layers increases, they are prone to issues such as vanishing or exploding gradients, leading to training instability and subsequent erratic fluctuations in loss values and recognition rates. To address this issue, a novel hybrid model, termed one-dimensional (1D) residual network with attention mechanism and bidirectional gated recurrent unit (BGRU) is developed for rotating machinery fault classification. First, a novel 1D residual network with optimized structure is constructed to obtain spatial features and mitigate the gradient vanishing or exploding. Second, the attention mechanism (AM) is designed to catch important impact characteristics for fault samples. Next, temporal features are mined through the BGRU. Finally, feature information is summarized through global average pooling, and the fully connected layer is utilized to output the final classification result for rotating machinery fault diagnosis. The developed technique which is tested on one set of planetary gear data and three different sets of bearing data, has achieved classification accuracy of 98.5%, 100%, 100%, and 100%, respectively. Compared with other methods, including CNN, CNN-BGRU, CNN-AM, and CNN with an AM-BGRU, the proposed technique has the highest recognition rate and stable diagnostic performance.

Research on roller monitoring technology based on distributed fiber optic sensing system

As one of the key components of the belt conveyor, the roller bears the task of supporting and rolling the conveyor belt, and monitoring its condition is very important. The traditional monitoring of the conveyor roller group adopts worker inspection, which has strong subjectivity. Monitoring using sensors necessitates the use of numerous sensors, which can pose wiring challenges. The use of inspection robots for monitoring results can be discontinuous, and their performance may be limited. This study proposes a fault diagnosis method for rollers based on a distributed fiber optic sensing system. By improving the traditional Isolation Forest (IForest), a framework called Incremental Majority Voting Isolation Forest (IMV-IForest) is proposed. By analyzing the optical signal, we extracted the variation patterns of roller faults over time and space, and analyzed the abnormal score distribution between fault data and normal data. Using the dataset collected on-site, we compared and analyzed IMV-IForest with the traditional IForest and the Extended Isolation Forest (E-iForest). The results indicate that the variation of the fault of the faulty roller with time and space can be used for early prediction of roller faults; determine an anomaly score threshold of 0.6; improved IForest have faster computation time and higher accuracy. Finally, to verify the effectiveness of the proposed scheme, a 3-month experiment was conducted on a 600 m long belt conveyor in a certain mine, and on-site monitoring results were obtained. By comparing with manual detection results, it was shown that the proposed method has high recognition rate for faulty idlers, with an accuracy rate of 97.92%, and can effectively diagnose faulty idlers.

Identification of rail cracks based on path graph features and SVM

In order to improve the accuracy of rail crack identification, a new method based on path graph feature and support vector machine is proposed. This method uses graph signal processing and graph theory to transform the magnetic flux leakage signal of rail crack, calculates the “time domain” and “frequency domain” statistics of the path graph signal, and effectively identifies rail cracks with different defect parameters by SVM classifier. The measured data verify the effectiveness of this method, which shows that the method of identifying rail cracks by using path graph features has higher accuracy and stability. The innovation of this method is that it draws on the idea of transform domain features to extract the graph domain features that can best represent the MFL signal. Compared with the 31 features used by the traditional method, this method only needs 22 features to achieve better recognition results and has shorter training time. For the recognition rate of 18 kinds of cracks, the average recognition rate of this method is more than 83.51%, and the highest recognition rate is 95.34%. Therefore, this study provides a new way for magnetic leakage analysis and treatment of rail crack detection, has important practical value, and provides beneficial enlightenment for further research in related fields.

Research on Intelligent System of Multimodal Deep Learning in Image Recognition

In this paper, a multi-scale image estimation method based on wavelet transform is proposed, which can effectively remove motion features from multiple videos. Then the autoencoder with sparsity limit is used to adjust the input signal to achieve effective compression. The effective features are extracted and the optimal unique vector is learned. The improved convolutional neural network is used to recognize weak moving objects. Experiments show that the algorithm can achieve high accuracy without large-scale learning samples, and the highest recognition rate is 99.36%. This algorithm has a great improvement over conventional algorithm.

HyperFace: A Deep Fusion Model for Hyperspectral Face Recognition.

Face recognition has been well studied under visible light and infrared (IR) in both intra-spectral and cross-spectral cases. However, how to fuse different light bands for face recognition, i.e., hyperspectral face recognition, is still an open research problem, which has the advantages of richer information retention and all-weather functionality over single-band face recognition. Thus, in this research, we revisit the hyperspectral recognition problem and provide a deep learning-based approach. A new fusion model (named HyperFace) is proposed to address this problem. The proposed model features a pre-fusion scheme, a Siamese encoder with bi-scope residual dense learning, a feedback-style decoder, and a recognition-oriented composite loss function. Experiments demonstrate that our method yields a much higher recognition rate than face recognition using only visible light or IR data. Moreover, our fusion model is shown to be superior to other general-purpose image fusion methods that are either traditional or deep learning-based, including state-of-the-art methods, in terms of both image quality and recognition performance.