Network Pattern Recognition Research Articles

Analyzing the Factors Influencing Time Delays in Korean Railroad Accidents

Railroads play a pivotal role in the Korean national economy, necessitating a thorough understanding of factors influencing accidents for effective mitigation strategies. Unlike prior research focused on accident frequency and severity, this study delves into the often-overlooked aspect of time delays resulting from railroad accidents. Analyzing 15 years of nationwide data (2008–2022), encompassing 3244 human-related and 3350 technical events, this research identifies key factors influencing delay likelihood and duration. Factors considered include event type, season, train type, location, operator size, person type involved, facility type, and causes. Despite an overall decrease in events, variable delay times highlight the need to comprehend specific contributing factors. To address excess zeros, the study employs a two-stage model and a zero-inflated negative binomial (ZINB) model, alongside artificial neural networks (ANNs) for non-linear pattern recognition. Human-related delays are influenced by event types, seasons, and passenger categories, exhibit nuanced impacts. Technical-related delays are influenced by incident types and facility involvement. Regarding model performance, the ANN models outperform regression-based models consistently in all cases. This study emphasizes the importance of considering both human and technical factors in predicting and understanding railroad accident delays, offering valuable insights for formulating strategies to mitigate service disruptions associated with these incidents.

Advanced Machine Learning Techniques in Gomoku: Strategy,Implementation, and Analysis

The strategic board game Gomoku has become a compelling domain for artificial intelligence (AI) research, particularly in developing and applying machine learning techniques. This paper comprehensively analyzes advanced machine learning strategies in Gomoku, focusing on logistic regression for board evaluation, neural networks for pattern recognition, and reinforcement learning for strategic gameplay. We discuss integrating these techniques in creating a sophisticated AI capable of high-level play and adaptability. Through this exploration, we highlight the potential of AI in strategic decision-making and its broader applications beyond board games.

A Comprehensive Review on Integrating Image Classification and Deep Learning for Theft Prevention through Emotion Detection

Abstract: This review delves into the synergistic integration of image classification and deep learning techniques to bolster theft prevention through the lens of emotion detection. The core of this exploration lies in grasping the fundamentals of image classification, specifically the application of Convolutional Neural Networks (CNNs) for intricate pattern recognition. The emphasis then shifts to the inclusion of emotion detection, a pioneering element that equips systems to discern emotional cues embedded in images. This encompasses scrutinizing facial expressions and contextual intricacies, affording a more intricate understanding of human behavior and potential security threats. The review showcases real-world case studies to exemplify the tangible impact of this integration, highlighting its effectiveness in curtailing false alarms and fortifying overall security infrastructures. However, amidst the promising outcomes, the review acknowledges challenges such as privacy concerns and the need for extensive datasets. In conclusion, the review envisions future directions, charting a course for refining and advancing the fusion of image classification and emotion detection. This amalgamation of cutting-edge technologies underscores a progressive paradigm in theft prevention, offering a comprehensive overview of its present state and the potential for further innovation.

Lulu Filterized Lin’s correlative Theil-Sen regression-based fully connected deep multilayer perceptive neural network for eye gaze pattern recognition

Gaze estimation is process finding the point of gaze on observe axis of eye. Gaze tracking schemes are mainly employed in HCI and study of visual scanning samples. Traditional tracking schemes usually need accurate personal calibration procedure to evaluate the particular eye metrics. In order to improve the accurate gaze estimation, Lulu Filterized Lin?s Correlative Theil-Sen Regression-based Fully Connected Deep Multilayer Perceptive Neural Network (LFLCTR-FCDMPNN) is designed for accurate gaze pattern identification through lesser time consumption. Fully Connected Deep Multilayer Perceptive NN contains input layer, three hidden layers, output layer. In input layer, number of gaze images is collected. Then using Lulu nonlinear smoothing filtering method is applied in initial hidden layer for removing noise as well as enhancing image quality. In second hidden layer, Polar coordinate system-based eye-gaze point estimation is performed. Finally, the Gaze Pattern matching is carried out in third hidden layer using Lin?s Concordance Correlative Theil-Sen regression. The estimated gaze points are organized at gaze plane to identify gaze patterns. Then pattern matching performed by Lin?s Concordance Correlation. In this way, the eye gaze patterns are correctly recognized at the output layer. Experimental evaluation is conducted to demonstrate performance analysis of LFLCTR-FCDMPNN technique through different metrics like gaze pattern recognition accuracy, gaze pattern recognition time, and false-positive rate with different number of eye images. Explained result illustrates which LFLCTR-FCDMPNN method improves the accuracy of gaze pattern recognition and decreases the time consumption than the conventional prediction methods. Using the Synthes Eyes dataset, it turned out that the FPR of the suggested LFLCTR-FCDMPNN was 63% higher than existing.

Diagnostic Method for Hydropower Plant Condition-based Maintenance combining Autoencoder with Clustering Algorithms

The French company EDF uses supervisory control and data acquisition systems in conjunction with a data management platform to monitor hydropower plant, allowing engineers and technicians to analyse the time-series collected. Depending on the strategic importance of the monitored hydropower plant, the number of time-series collected can vary greatly making it difficult to generate valuable information from the extracted data. In an attempt to provide an answer to this particular problem, a condition detection and diagnosis method combining clustering algorithms and autoencoder neural networks for pattern recognition has been developed and is presented in this paper.First, a dimension reduction algorithm is used to create a 2- or 3-dimensional projection that allows the users to identify unsuspected relationships between datapoints.Then, a collection of clustering algorithms regroups the datapoints into clusters. For each identified cluster, an autoencoder neural network is trained on the corresponding dataset. The aim is to measure the reconstruction error between each autoencoder model and the measured values, thus creating a proximity index for each state discovered during the clustering stage.

Application of Neural Networks for Pattern Recognition

This article details the process of creating and training a neural network for the task of classifying dog and cat images using TensorFlow and the MobileNetV2 architecture. Data preparation and preprocessing, including image resizing and normalization, are described. Details of the integration of the pre-trained MobileNetV2 model are given, the process of pre-training the model on specific data is demonstrated, as well as methods of model construction and optimization, including the addition and tuning of additional layers. Special attention is paid to practical application of the trained model for classification of new images, including loading, processing, prediction and visualization of results.

Underwater target material classification based on feature extraction of finite element simulated signal

Feature extraction is crucial for underwater target classification and recognition, and this paper aims to propose a scheme for accurately classifying the material features of underwater targets. Finite element simulation software COMSOL is used to obtain echo signals, and then Auto Regressive (AR) coefficient, cepstrum feature, spectral peak and its frequency of Smoothing Pseudo Wigner-Ville Distribution (SPWVD) of echo signals are extracted as the input of the pattern recognition network. The classification results show that the proposed scheme is able to effectively differentiate among four types of materials: metal, brick, plastic and rubber. The scheme also shows good robustness for the target size and shape, and the proposed scheme has the potential to be applied to the practical underwater target classification.

Artificial visual‐tactile perception array for enhanced memory and neuromorphic computations

AbstractThe emulation of human multisensory functions to construct artificial perception systems is an intriguing challenge for developing humanoid robotics and cross‐modal human–machine interfaces. Inspired by human multisensory signal generation and neuroplasticity‐based signal processing, here, an artificial perceptual neuro array with visual‐tactile sensing, processing, learning, and memory is demonstrated. The neuromorphic bimodal perception array compactly combines an artificial photoelectric synapse network and an integrated mechanoluminescent layer, endowing individual and synergistic plastic modulation of optical and mechanical information, including short‐term memory, long‐term memory, paired pulse facilitation, and “learning‐experience” behavior. Sequential or superimposed visual and tactile stimuli inputs can efficiently simulate the associative learning process of “Pavlov's dog”. The fusion of visual and tactile modulation enables enhanced memory of the stimulation image during the learning process. A machine‐learning algorithm is coupled with an artificial neural network for pattern recognition, achieving a recognition accuracy of 70% for bimodal training, which is higher than that obtained by unimodal training. In addition, the artificial perceptual neuron has a low energy consumption of ∼20 pJ. With its mechanical compliance and simple architecture, the neuromorphic bimodal perception array has promising applications in large‐scale cross‐modal interactions and high‐throughput intelligent perceptions.image

FPRnet: A lightweight multi-domain multi-stream network for complex horizontal oil-water two-phase flow pattern recognition

The complicated spatio-temporal structure of horizontal oil-water flows renders the identification of flow patterns exceedingly challenging, owing to the presence of interfacial waves and droplet entrainment. In this study, a multi-domain fusion feature (MFF) as flow pattern descriptors (FPD) and a lightweight multi-stream networks (LMN) as flow pattern classifier (FPC) are proposed to identify the flow patterns of horizontal oil-water two-phase flow. MFF can map time series into different domains and compactly encode them into higher dimensional expressions, thus displaying implicit flow pattern signals incorporating nonlinear, time-frequency, and probabilistic transition domain features. Finally, the comparison of LMN, variational and other networks is evaluated on the horizontal oil-water flows pattern identification task. Extensive comparison experiments show that the time-frequency feature exhibits optimal characterization for the D O/W flow pattern, the URP feature is most accurately characterized for the D W/O and D O/W flow patterns, and the MTF feature exhibits superior characterization for the remaining three flow patterns. The 98.40 % recognition accuracy demonstrates that MFF has a good representation of flow patterns and the proposed LMN outperforms existing classifiers in flow pattern recognition.

The KI-ASIC Dataset

We present a novel dataset captured from a BMW X5 test carrier within the German research project KI-ASIC for use in radar sensor development and autonomous driving research. Our work aims at providing a blueprint for the process of creating labeled datasets for the development of neural networks for pattern recognition in radar data in the automotive environment. With a variety of different sensor types such as wide angle color cameras, a high-resolution color stereo camera, an Ouster OS1-64 laser scanner and three novel Infineon radar sensors, we recorded over 100,000 scenes of real traffic scenarios as well as defined test scenarios with a frequency of 10 Hz. The scenarios in real traffic contain inner-city situations, but also scenes from rural areas with static and dynamic objects. Besides, the defined test scenarios are based on the NCAP scenarios and focus mostly on turning, overtaking and follow-up maneuvers. The data from the different sensors is calibrated, synchronized and timestamped including raw and rectified information. Our dataset also contains labels for all detected objects from a defined class list with distance and angle properties. The content of the paper aims at the description of the recording test carrier, the format of the provided sensor data and the structure of the overall dataset

An EEG-based cross-subject interpretable CNN for game player expertise level classification

Electroencephalogram (EEG) signals have been demonstrated to be an effective method for game player expertise level classification, as it can reflect the activity state of the player’s brain during the game task. Although many efforts have been made to classify the expertise level of game players based on EEG signals, existing methods still need to be improved in identifying common brain patterns across different subjects. In this article, we propose a Spatiotemporal-Based Brain Pattern Recognition Network (BPR-STNet), which uses depthwise separable convolution to extract the spatiotemporal features of EEG signals and incorporates Gradient-weighted Class Activation Mapping (Grad-CAM) interpretable technology to explore common brain patterns among different subjects. To evaluate the model’s performance, we use non-invasive wearable devices to collect EEG signals from 19 subjects during the game process. The results of the leave-one-out cross-subject experiments demonstrate that the average accuracy of the model’s game player expertise level classification is 82%–86.32% in the five frequency bands (δ, θ, α, β, γ). Among them, the γ frequency band has the highest accuracy of 86.32%. More importantly, the average accuracy of our method is 1.56%–5.79% higher than the state-of-the-art deep learning methods. Moreover, interpretable results indicate the model can learn biologically significant features from EEG frequency bands. Overall, this study can offer a new idea of using deep learning to explore common brain patterns and features among different subjects at the same level. The BPR-STNet code is available at https://github.com/L000077/BPR-STNet.

Sound processing in the cricket brain: evidence for a pulse duration filter.

The auditory system of female crickets allows them to specifically recognize and approach the species-specific male calling song, defined by sound pulses and silent intervals. Auditory brain neurons form a delay-line and coincidence detector network tuned to the pulse period of the male song. We analyzed the impact of changes in pulse duration on the behavior and the responses of the auditory neurons and the network. We confirm that the ascending neuron AN1 and the local neuron LN2 copy the temporal structure of the song. During ongoing long sound pulses, the delay-line neuron LN5 shows additional rebound responses and the coincidence detector neuron LN3 can generate additional bursts of activity, indicating that these may be driven by intrinsic oscillations of the network. Moreover, the response of the feature detector neuron LN4 is shaped by a combination of inhibitory and excitatory synaptic inputs, and LN4 responds even to long sound pulses with a short depolarization and burst of spikes, like to a sound pulse of natural duration. This response property of LN4 indicates a selective auditory pulse duration filter mechanism of the pattern recognition network, which is tuned to the duration of natural pulses. Comparing the tuning of the phonotactic behavior with the tuning of the local auditory brain neurons to the same test patterns, we find no evidence that a modulation of the phonotactic behavior is reflected at the level of the feature detector neurons. This rather suggests that steering to nonattractive pulse patterns is organized at the thoracic level.NEW & NOTEWORTHY Pulse period selectivity has been reported for the cricket delay-line and coincidence detector network, whereas pulse duration selectivity is evident from behavioral tests. Pulses of increasing duration elicit responses in the pattern recognition neurons, which do not parallel the behavioral responses and indicate additional processing mechanisms. Long sound pulses elicit rhythmic rebound activity and additional bursts, whereas the feature detector neuron reveals a pulse duration filter, expanding our understanding of the pattern recognition process.

A network classification method based on density time evolution patterns extracted from network automata

Network modeling has proven to be an efficient tool for many interdisciplinary areas, including social, biological, transportation, and various other complex real-world systems. In addition, cellular automata (CA) are a formalism that has received significant attention in recent decades as a model for investigating patterns in the dynamic spatio-temporal behavior of these systems, based on local rules. Some studies investigate the use of cellular automata to analyze the dynamic behavior of networks and refer to them as network automata (NA). Recently, it has been demonstrated that NA is effective for network classification, as it employs a Time-Evolution Pattern (TEP) for feature extraction. However, the TEPs investigated in previous studies consist of binary values (states) that do not capture the intrinsic details of the analyzed network. Therefore, in this work, we propose alternative sources of information that can be used as descriptors for the classification task, which we refer as Density Time-Evolution Pattern (D-TEP) and State Density Time-Evolution Pattern (SD-TEP). We examine the density of alive neighbors of each node, which is a continuous value, and compute feature vectors based on histograms of TEPs. Our results demonstrate significant improvement over previous studies on five synthetic network datasets, as well as seven real datasets. Our proposed method is not only a promising approach for pattern recognition in networks, but also shows considerable potential for other types of data that can be transformed into network.

A Novel Event-Driven Spiking Convolutional Neural Network for Electromyography Pattern Recognition.

Electromyography (EMG) pattern recognition is an important technology for prosthesis control and human-computer interaction etc. However, the practical application of EMG pattern recognition is hampered by poor accuracy and robustness due to electrode shift caused by repeated wearing of the signal acquisition device. Moreover, the user's acceptability is low due to the heavy training burden, which is caused by the need for a large amount of training data by traditional methods. In order to explore the advantage of spiking neural network (SNN) in solving the poor robustness and heavy training burden problems in EMG pattern recognition, a spiking convolutional neural network (SCNN) composed of cyclic convolutional neural network (CNN) and fully connected modules is proposed and implemented in this study. High density surface electromyography (HD-sEMG) signals collected from 6 gestures of 10 subjects at 6 electrode positions are taken as the research object. Compared to CNN with the same structure, CNN-Long Short Term Memory (CNN-LSTM), linear kernel linear discriminant analysis classifier (LDA) and spiking multilayer perceptron (Spiking MLP), the accuracy of SCNN is 50.69%, 33.92%, 32.94% and 9.41% higher in the small sample training experiment, 6.50%, 4.23%, 28.73%, and 2.57% higher in the electrode shifts experiment respectively. In addition, the power consumption of SCNN is about 1/93 of CNN. The advantages of the proposed framework in alleviating user training burden, mitigating the adverse effect of electrode shifts and reducing power consumption make it very meaningful for promoting the development of user-friendly real-time myoelectric control system.

A Novel Back-Propagation Neural Network for Intelligent Cyber-Physical Systems for Wireless Communications

Wireless sensor networks, which play a significant role in monitoring complex environments that change rapidly over time, were used in the Artificial Intelligence method. External factors or the device designers themselves are both responsible for this complex behavior. Sensor networks often use machine learning techniques to adapt to such conditions, eliminating the need for excessive redesign. Cyber-physical systems (CPS) appeared as the promising option for improving physical-virtual interactions. The quality of the system containing processing information is primarily determined by the system function. There are many benefits obtained while combining Artificial Intelligence (AI) and Cyber-Physical Systems (CPSs) in buildings. In CPS-based indoor environment has various design schemes containing measurement and intelligent buildings in the control system consisting of detection, tracking, execution, and communication modules. The Multi-Agent System (MAS) is the smallest control unit that simulates among neurons and it flexibly provides the information. To mimic the interactions between human neurons, multi-agents are used. In this paper, the CPS’s information world is built on the fundamental principle of granular formal concepts and the theory of granular computing is investigated. The calculation module is used by Back-Propagation Neural Network (BPNN) for pattern recognition and classification by environmental information. Various parameters namely the normalized root mean square error, peak signal-to-noise ratio, mean square error, and the mean absolute error are chosen as the objective assessment criteria to assess the benefits of the proposed method and the effectiveness of the proposed system is proven.

A Cascade Network for Pattern Recognition Based on Radar Signal Characteristics in Noisy Environments

Target recognition mainly focuses on three approaches: optical-image-based, echo-detection-based, and passive signal-analysis-based methods. Among them, the passive signal-based method is closely integrated with practical applications due to its strong environmental adaptability. Based on passive radar signal analysis, we design an “end-to-end” model that cascades a noise estimation network with a recognition network to identify working modes in noisy environments. The noise estimation network is implemented based on U-Net, which adopts a method of feature extraction and reconstruction to adaptively estimate the noise mapping level of the sample, which can help the recognition network to reduce noise interference. Focusing on the characteristics of radar signals, the recognition network is realized based on the multi-scale convolutional attention network (MSCANet). Firstly, deep group convolution is used to isolate the channel interaction in the shallow network. Then, through the multi-scale convolution module, the finer-grained features of the signal are extracted without increasing the complexity of the model. Finally, the self-attention mechanism is used to suppress the influence of low-correlation and negative-correlation channels and spaces. This method overcomes the problem of the conventional method being seriously disturbed by noise. We validated the proposed method in 81 kinds of noise environment, achieving an average accuracy of 94.65%. Additionally, we discussed the performance of six machine learning algorithms and four deep learning algorithms. Compared to these methods, the proposed MSCANet achieved an accuracy improvement of approximately 17%. Our method demonstrates better generalization and robustness.

Urban traffic congestion alleviation system based on millimeter wave radar and improved probabilistic neural network

AbstractThe millimeter‐wave radar sensor is widely used for urban traffic surveillance because of its weather resistance and high detection accuracy. Methods such as fuzzy theory, pattern recognition, and artificial neural networks have been integrated into the research of traffic state discrimination. However, research on systematically describing the fusion of sensors and traffic state discrimination algorithms to alleviate urban road congestion is still lacking, especially based on millimeter‐wave radar. Thus, the authors propose an urban traffic congestion alleviation system framework. First, the design and deployment of the millimeter‐wave radar system, including waveforms, signal processing flow, and target tracking, are demonstrated to achieve vehicle information acquisition and output. Then, the appropriate traffic parameters are obtained by analysing traffic state influencing factors and the radar data characteristics. Finally, a traffic conditions identification algorithm combining spectral clustering and neural network algorithm is presented to realise road congestion level classification. The system is applied to real urban intersections rather than simulation or approximate real simulation. According to the current road congestion level, regulate the traffic light state to achieve road vehicle driving command. Experiments show that the proposed system can effectively reduce road congestion by 20% compared to the current fixed traffic light system.

Brain-inspired nanophotonic spike computing: challenges and prospects

Nanophotonic spiking neural networks (SNNs) based on neuron-like excitable subwavelength (submicrometre) devices are of key importance for realizing brain-inspired, power-efficient artificial intelligence (AI) systems with high degree of parallelism and energy efficiency. Despite significant advances in neuromorphic photonics, compact and efficient nanophotonic elements for spiking signal emission and detection, as required for spike-based computation, remain largely unexplored. In this invited perspective, we outline the main challenges, early achievements, and opportunities toward a key-enabling photonic neuro-architecture using III–V/Si integrated spiking nodes based on nanoscale resonant tunnelling diodes (nanoRTDs) with folded negative differential resistance. We utilize nanoRTDs as nonlinear artificial neurons capable of spiking at high-speeds. We discuss the prospects for monolithic integration of nanoRTDs with nanoscale light-emitting diodes and nanolaser diodes, and nanophotodetectors to realize neuron emitter and receiver spiking nodes, respectively. Such layout would have a small footprint, fast operation, and low power consumption, all key requirements for efficient nano-optoelectronic spiking operation. We discuss how silicon photonics interconnects, integrated photorefractive interconnects, and 3D waveguide polymeric interconnections can be used for interconnecting the emitter-receiver spiking photonic neural nodes. Finally, using numerical simulations of artificial neuron models, we present spike-based spatio-temporal learning methods for applications in relevant AI-based functional tasks, such as image pattern recognition, edge detection, and SNNs for inference and learning. Future developments in neuromorphic spiking photonic nanocircuits, as outlined here, will significantly boost the processing and transmission capabilities of next-generation nanophotonic spike-based neuromorphic architectures for energy-efficient AI applications. This perspective paper is a result of the European Union funded research project ChipAI in the frame of the Horizon 2020 Future and Emerging Technologies Open programme.

Electroencephalogram signal classification based on Fourier transform and Pattern Recognition Network for epilepsy diagnosis

Epilepsy is a central nervous system (CNS) disorder that affects nerve cells in the brain and produces seizures in which consciousness is lost. People with epilepsy have frequent seizures as a result of increased brain electrical activity, which disrupts the message system between brain cells, making epilepsy a serious condition that must be treated. Correct diagnosis of this disease can save the patient from death and complications caused by the disease. The recording of electroencephalogram (EEG) signals is an effective method for determining the electrical activity of the brain and learning a great deal about how the brain function. The Fast Fourier Transform (FFT), the Discrete Wavelet Transform (DWT), and a pattern recognition network are used in this study to offer a method for analyzing EEG signals. This study aimed to provide a multi-step algorithm for extracting signal features and diagnosing epilepsy. Wavelet transform is used to remove signal noise. For this purpose, the optimal wavelet mother (Dabichiz 8) was used. Then, signal features are extracted using the Fourier transform, and generate the EEG matrix. The features are regarded as the Pattern Recognition Network’s input. There were 5% of test data and 80% of training data. 15% of the data were left over for validation. The architecture was thought to consist of one input layer (14 neurons = number of selected features), one hidden layer (6 neurons), and two output layers (2 neurons = number of sleep stages). The result of this output was compared with other classifiers, such as multilayer perceptron (MLP) neural network. The results show that using a Pattern Recognition Network can classify the features with 92.5% accuracy. Using this method can process the signal with high accuracy while being simple. In fact, a new method based on the architecture of a new network was presented for AC signal processing. Hence, the proposed method can diagnose epilepsy with high accuracy while removing signal noise.

A Cloud-Based Urban Monitoring System by Using a Quadcopter and Intelligent Learning Techniques

The application of quadcopter and intelligent learning techniques in urban monitoring systems can improve the flexibility and efficiency features. This paper proposes a cloud-based urban monitoring system that uses deep learning, fuzzy system, image processing, pattern recognition, and Bayesian network. Main objectives of this system are to monitor the climate status, temperature, humidity, and smoke as well as to detect the fire occurrences based on the above intelligent techniques. The quadcopter transmits sensing data of the temperature, humidity, and smoke sensors, geographical coordinates, image frames, and videos to a control station via RF communications. In the control station side, the monitoring capabilities are designed by graphical tools to show urban areas with RGB colors according to the pre-determined data ranges. The evaluation process illustrates simulation results of the deep neural network applied to climate status, effects of the sensors’ data changes on climate status. An illustrative example is used to draw the simulated area by RGB colors. Furthermore, circuit of the quadcopter side is designed using electric devices.