Dataset Of Signals Research Articles

Ml assisted techniques in power side channel analysis for trojan classification

Hardware Trojans (HTs) are one of the emerging malicious hardware modification attacks that have become a critical threat to the integrity, reliability, security, and trustworthiness of integrated circuits (ICs) applications. These deliberately hidden malicious entities can be inserted into the IC during manufacturing or design, potentially leading to the leakage of secret information or the deactivation or destruction of the entire system that relies on the IC hardware chips. Localizing and detecting hardware Trojans is becoming increasingly challenging as these threats are deeply embedded and electronic systems continue to grow in complexity. Traditional methods, including physical inspections and functional testing, are increasingly inadequate. They are limited in scope and often fall short when confronted with the advanced designs of modern hardware Trojans.This study aims to improve the identification of hardware Trojans (HTs) in integrated circuits by using advanced machine learning algorithms with a unique dataset of power side-channel signals. Various machine learning techniques, such as Support Vector Machines (SVM), neural networks, and decision trees, are applied to classify and identify HTs accurately. This method combines comprehensive feature extraction and model validation to provide high accuracy and reliability in HT identification. This research makes significant contributions to cybersecurity by providing improved techniques for detecting minor anomalies associated with HTs using advanced machine learning algorithms. The results show potential advances in securing electronic systems against these hidden threats, highlighting the practical significance and necessity of our work in maintaining a wide range of applications from consumer electronics to national infrastructure.

Guided Waves in Composite Overwrapped Pressure Vessels and Considerations for Sensor Placement Towards Structural Health Monitoring – An Experimental Study

Abstract The utilization of composite overwrapped pressure vessels (COPVs) to store hydrogen, especially at high pressures, is gaining more popularity due to their lightweight design and high storage density, offering significant economic advantages. However, the presence of material defects or fatigue can lead to critical failures, requiring an innovative and robust approach to ensure safe operation and system integrity. Developing a continuous structural health monitoring (SHM) system for COPVs can provide comprehensive real-time information about their condition, facilitating a shift away from periodic inspections. This study scrutinizes the behavior of guided waves (GWs) within COPVs to design a sensor network for damage detection and localization. First, the dispersive and multimodal propagation behavior of GWs is experimentally investigated. Subsequently, important parameters for the network design are derived and finally a sensor network consisting of fifteen piezoelectric transducers is designed to cover the entire cylindrical area. The effectiveness is then evaluated experimentally by placing artificial defects on the surface of the COPV. The multi-layered dataset of GW signals was analyzed using both commonly used ultrasonic features (e.g., amplitude, frequency, time of flight) as well as statistical features (kurtosis, skewness, variance, etc.). These features were utilized to compute a damage index, and the effectiveness of the detection performance was assessed using receiver operating characteristic curves. It can be seen that some features are more sensitive and robust under varying experimental conditions. The results show that ultrasonic GW SHM system is a promising solution for damage detection and localization in COPVs.

Enhancement of structural and functional photoacoustic imaging based on a reference-inputted convolutional neural network.

Photoacoustic microscopy has demonstrated outstanding performance in high-resolution functional imaging. However, in the process of photoacoustic imaging, the photoacoustic signals will be polluted by inevitable background noise. Besides, the image quality is compromised due to the biosafety limitation of the laser. The conventional approach to improving image quality, such as increasing laser pulse energy or multiple-times averaging, could result in more health risks and motion artifacts for high exposures to the laser. To overcome this challenge of biosafety and compromised image quality, we propose a reference-inputted convolutional neural network (Ri-Net). The network is trained using the photoacoustic signal and noise datasets from phantom experiments. Evaluation of the trained neural network demonstrates significant signal improvement. Human cuticle microvasculature imaging experiments are also conducted to further assess the performance and practicality of our network. The quantitative results show that we achieved a 2.6-fold improvement in image contrast and a 9.6 dB increase in signal-to-noise ratio. Finally, we apply our network, trained on single-wavelength data, to multi-wavelength functional imaging. The functional imaging of the mouse ear demonstrates the robustness of our method and the potential to capture the oxygen saturation of microvasculature. The Ri-Net enhances photoacoustic microscopy imaging, allowing for more efficient microcirculation assessments in a clinical setting.

Learning Motor Cues in Brain-Muscle Modulation.

Current studies for brain-muscle modulation often analyze selected properties in electrophysiological signals, leading to a partial understanding. This article proposes a cross-modal generative model that converts brain activities measured by electroencephalography (EEG) to corresponding muscular responses recorded by electromyography (EMG). Examining the generation process in the model highlights how the motor cue, representing implicit motor information hidden within brain activities, modulates the interaction between brain and muscle systems. The proposed model employs a two-stage generation process to bridge the semantic gap in cross-modal signals. Initially, the shared movement-related information between EEG and EMG signals is extracted using a contrastive learning framework. These shared representations act as conditional vectors in the subsequent EMG generation stage based on generative adversarial networks (GANs). Experiments on a self-collected multimodal electrophysiological signal data set show the algorithm's superiority over existing time series generative methods in cross-modal EMG generation. Further insights derived from the model's inference process underscore the brain's strategy for muscle control during movements. This research provides a data-driven approach for the neuroscience community, offering a comprehensive perspective of brain-muscular modulation.

Research on automatic modulation recognition of shortwave signals in few‐shot scenarios based on knowledge distillation

AbstractAutomatic modulation recognition plays an important role in wireless communication and radio regulation. Existing deep learning‐based automatic modulation recognition techniques perform well with large datasets and high computational power but require significant resources for labelled data and complex pre‐processing. This paper proposes a multi‐information fusion method for few‐shot modulation recognition, which involves converting I/Q signals into A/P signals and training with a combination of VGG and LSTM network models. An ensemble knowledge distillation (EKD) approach is employed to streamline the network model, meeting the demands for deploying neural network models on compact devices. Experimental results demonstrate that using only 1% of the shortwave modulation signal dataset as the training set, the proposed model achieves an average classification accuracy of 71.08% under all signal‐to‐noise ratios, surpassing the currently popular deep learning models. Moreover, two small‐scale networks, MobileNetV3 and convolutional neural network are trained, through EKD. Compared to the teacher network, the floating‐point operations of the distilled models are reduced by 99.8% and 99.7%, respectively, and the average prediction accuracy only decreases by 16.05% and 8.09%. The lightweight, few‐shot networks designed in this study for shortwave modulation signals aim to achieve fast and accurate modulation recognition on compact devices.

A Novel Linear Discriminant Analysis Based Classification of R-peaks in Different ECG Signal Datasets

A Novel Linear Discriminant Analysis Based Classification of R-peaks in Different ECG Signal Datasets

Assessing the documentation of publicly available medical image and signal datasets and their impact on bias using the BEAMRAD tool

AbstractMedical datasets are vital for advancing Artificial Intelligence (AI) in healthcare. Yet biases in these datasets on which deep-learning models are trained can compromise reliability. This study investigates biases stemming from dataset-creation practices. Drawing on existing guidelines, we first developed a BEAMRAD tool to assess the documentation of public Magnetic Resonance Imaging (MRI); Color Fundus Photography (CFP), and Electrocardiogram (ECG) datasets. In doing so, we provide an overview of the biases that may emerge due to inadequate dataset documentation. Second, we examine the current state of documentation for public medical images and signal data. Our research reveals that there is substantial variance in the documentation of image and signal datasets, even though guidelines have been developed in medical imaging. This indicates that dataset documentation is subject to individual discretionary decisions. Furthermore, we find that aspects such as hardware and data acquisition details are commonly documented, while information regarding data annotation practices, annotation error quantification, or data limitations are not consistently reported. This risks having considerable implications for the abilities of data users to detect potential sources of bias through these respective aspects and develop reliable and robust models that can be adapted for clinical practice.

Estimation of the High-Frequency Feature Slope in Gravitational Wave Signals from Core Collapse Supernovae Using Machine Learning

We conducted an in-depth exploration of the use of different machine learning (ML) for regression algorithms, including Linear, Ridge, LASSO, Bayesian Ridge, Decision Tree, and a variety of Deep Neural Network (DNN) architectures, to estimate the slope of the high-frequency feature (HFF), a prominent emergent feature found in the gravitational wave (GW) signals of core collapse supernovae (CCSN). We created a data set of CCSN GW signals generated by an analytical model that mimics the characteristics of the signals obtained from numerical simulations, particularly the HFF. This enabled us to simulate a wide range of HFF slope values and analyze their properties. We opted to employ ML for regression techniques, particularly a supervised learning approach, to analyze the data set due to the parameter chosen for estimating the slope of the HFF. This type of architecture is ideal for this purpose as it can detect the connections between input and output data. In addition, it is suitable for handling high-dimensional input data and produces efficient results with low computational cost. We evaluated the efficiency and performance of the ML algorithms using a set of metrics to measure their ability to accurately predict the HFF slope within the data set. The results showed that a DNN algorithm for regression exhibits the highest accuracy in estimating the slope of the HFF.

A One-Dimensional Depthwise Separable Convolutional Neural Network for Bearing Fault Diagnosis Implemented on FPGA.

This paper presents a hardware implementation of a one-dimensional convolutional neural network using depthwise separable convolution (DSC) on the VC707 FPGA development board. The design processes the one-dimensional rolling bearing current signal dataset provided by Paderborn University (PU), employing minimal preprocessing to maximize the comprehensiveness of feature extraction. To address the high parameter demands commonly associated with convolutional neural networks (CNNs), the model incorporates DSC, significantly reducing computational complexity and parameter load. Additionally, the DoReFa-Net quantization method is applied to compress network parameters and activation function outputs, thereby minimizing memory usage. The quantized DSC model requires approximately 22 KB of storage and performs 1,203,128 floating-point operations in total. The implementation achieves a power consumption of 527 mW at a clock frequency of 50 MHz, while delivering a fault diagnosis accuracy of 96.12%.

Deep Neural Networks for Estimating Regularization Parameter in Sparse Time–Frequency Reconstruction

Time–frequency distributions (TFDs) are crucial for analyzing non-stationary signals. Compressive sensing (CS) in the ambiguity domain offers an approach for TFD reconstruction with high performance, but selecting the optimal regularization parameter for various signals remains challenging. Traditional methods for parameter selection, including manual and experimental approaches, as well as existing optimization procedures, can be imprecise and time-consuming. This study introduces a novel approach using deep neural networks (DNNs) to predict regularization parameters based on Wigner–Ville distributions (WVDs). The proposed DNN is trained on a comprehensive dataset of synthetic signals featuring multiple linear and quadratic frequency-modulated components, with variations in component amplitudes and random positions, ensuring wide applicability and robustness. By utilizing DNNs, end-users need only provide the signal’s WVD, eliminating the need for manual parameter selection and lengthy optimization procedures. Comparisons between the reconstructed TFDs using the proposed DNN-based approach and existing optimization methods highlight significant improvements in both reconstruction performance and execution time. The effectiveness of this methodology is validated on noisy synthetic and real-world signals, emphasizing the potential of DNNs to automate regularization parameter determination for CS-based TFD reconstruction in diverse signal environments.

A Novel Improved Method for Prediction of Heart Disease using ECG Hybrid 0F PTB-ECG and MIT-BIH Signal Dataset

Heart disease is the leading cause of death worldwide, making early detection critical. Various diagnostic methods, including clinical tests, CT, MRI, ECG, and impedance cardiography, are commonly used to detect heart disease. However, traditional coronary artery disease (CAD) detection methods using ECG data face challenges due to the time-series nature of ECG signals, which complicates handling multiple classes. To address this, the study proposes a deep learning-based approach that enhances CAD detection accuracy by integrating two models Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) with a hybrid dataset combining PTB-ECG and MIT-BIH data. This hybrid dataset consists of two target classes: normal (0) and abnormal (1), created by merging all MIT-BIH classes with the PTB-ECG normal class as “0” and abnormal samples from PTB-ECG as “1”. Pre-processing was performed using Gaussian distribution for normalization, standardization, and outlier removal. The study applied four classification approaches: CNN, CNN+LSTM, CNN with SMOTE-balanced data, and CNN+LSTM with SMOTE-balanced data. Results indicate that CNN with SMOTE-balanced data achieved the best performance, with training metrics of 0.9998 accuracy, 1.00 precision, 1.00 recall, and 1.00 F1-score for both classes. Testing results using CNN+SMOTE reached 0.9991 accuracy, 1.00 precision, 1.00 recall, and 1.00 F1-score. The model surpasses state-of-the-art studies, which achieved 0.992 accuracy and F1-score of 0.986 on PTB-ECG and MIT-BIH datasets, respectively. This study demonstrates that combining CNN with SMOTE on a hybrid dataset can significantly improve CAD detection accuracy.

Cardioish: Lead-Based Feature Extraction for ECG Signals.

Background: Electrocardiography (ECG) signals are commonly used to detect cardiac disorders, with 12-lead ECGs being the standard method for acquiring these signals. The primary objective of this research is to propose a new feature engineering model that achieves both high classification accuracy and explainable results using ECG signals. To this end, a symbolic language, named Cardioish, has been introduced. Methods: In this research, two publicly available datasets were used: (i) a mental disorder classification dataset and (ii) a myocardial infarction (MI) dataset. These datasets contain ECG beats and include 4 and 11 classes, respectively. To obtain explainable results from these ECG signal datasets, a new explainable feature engineering (XFE) model has been proposed. The Cardioish-based XFE model consists of four main phases: (i) lead transformation and transition table feature extraction, (ii) iterative neighborhood component analysis (INCA) for feature selection, (iii) classification, and (iv) explainable results generation using the recommended Cardioish. In the feature extraction phase, the lead transformer converts ECG signals into lead indexes. To extract features from the transformed signals, a transition table-based feature extractor is applied, resulting in 144 features (12 × 12) from each ECG signal. In the feature selection phase, INCA is used to select the most informative features from the 144 generated, which are then classified using the k-nearest neighbors (kNN) classifier. The final phase is the explainable artificial intelligence (XAI) phase. In this phase, Cardioish symbols are created, forming a Cardioish sentence. By analyzing the extracted sentence, XAI results are obtained. Additionally, these results can be integrated into connectome theory for applications in cardiology. Results: The presented Cardioish-based XFE model achieved over 99% classification accuracy on both datasets. Moreover, the XAI results related to these disorders have been presented in this research. Conclusions: The recommended Cardioish-based XFE model achieved high classification performance for both datasets and provided explainable results. In this regard, our proposal paves a new way for ECG classification and interpretation.

An accurate semi-empirical model for PMT pulse signal analysis.

The energy information of pulse signals is significantly important for applications such as computed tomography (CT), positron emission tomography (PET), and research on defects in condensed matter. Time-over-threshold (TOT) and multi-voltage threshold (MVT) are commonly used digitization methods in sampling pulse signal. However, both approaches rely on a mathematical model of the pulse signal to derive energy information. This study proposes a semi-empirical mathematical model for pulse signals formation process in scintillation crystal-coupled photomultiplier tube(PMT) probes, by utilizing the CR-RC shaping method. This mathematical model accurately describes output of the PMT pulse signals. This study analyzes a substantial dataset of pulse signals, comparing the performance of the newly designed mathematical model with that of the double exponential function in terms of their ability to fit pulse signals. The results indicate that the mathematical model developed herein achieves an average R2 of 0.9255, significantly surpassing the 0.9155 of the double exponential function, thereby demonstrating its superior fitting efficacy.

ChMinMaxPat: Investigations on Violence and Stress Detection Using EEG Signals.

Electroencephalography (EEG) signals, often termed the letters of the brain, are one of the most cost-effective methods for gathering valuable information about brain activity. This study presents a new explainable feature engineering (XFE) model designed to classify EEG data for violence detection. The primary objective is to assess the classification capability of the proposed XFE model, which uses a next-generation feature extractor, and to obtain interpretable findings for EEG-based violence and stress detection. In this research, two distinct EEG signal datasets were used to obtain classification and explainable results. The recommended XFE model utilizes a channel-based minimum and maximum pattern (ChMinMaxPat) feature extraction function, which generates 15 distinct feature vectors from EEG data. Cumulative weight-based neighborhood component analysis (CWNCA) is employed to select the most informative features from these vectors. Classification is performed by applying an iterative and ensemble t-algorithm-based k-nearest neighbors (tkNN) classifier to each feature vector. Information fusion is achieved through iterative majority voting (IMV), which consolidates the 15 tkNN classification results. Finally, the Directed Lobish (DLob) symbolic language generates interpretable outputs by leveraging the identities of the selected features. Together, the tkNN classifier, IMV-based information fusion, and DLob-based explainable feature extraction transform the model into a self-organizing explainable feature engineering (SOXFE) framework. The ChMinMaxPat-based model achieved over 70% accuracy on both datasets with leave-one-record-out (LORO) cross-validation (CV) and over 90% accuracy with 10-fold CV. For each dataset, 15 DLob strings were generated, providing explainable outputs based on these symbolic representations. The ChMinMaxPat-based SOXFE model demonstrates high classification accuracy and interpretability in detecting violence and stress from EEG signals. This model contributes to both feature engineering and neuroscience by enabling explainable EEG classification, underscoring the potential importance of EEG analysis in clinical and forensic applications.

Intelligent Fault Diagnosis of Inter-Turn Short Circuit Faults in PMSMs for Agricultural Machinery Based on Data Fusion and Bayesian Optimization

The permanent magnet synchronous motor (PMSM) plays an important role in the power system of agricultural machinery. Inter-turn short circuit (ITSC) faults are among the most common failures in PMSMs, and early diagnosis of these faults is crucial for enhancing the safety and reliability of motor operation. In this article, a multi-source data-fusion algorithm based on convolutional neural networks (CNNs) has been proposed for the early fault diagnosis of ITSCs. The contributions of this paper can be summarized in three main aspects. Firstly, synchronizing data from different signals extracted by different devices presents a significant challenge. To address this, a signal synchronization method based on maximum cross-correlation is proposed to construct a synchronized dataset of current and vibration signals. Secondly, applying a traditional CNN to the data fusion of different signals is challenging. To solve this problem, a multi-stream high-level feature fusion algorithm based on a channel attention mechanism is proposed. Thirdly, to tackle the issue of hyperparameter tuning in deep learning models, a hyperparameter optimization method based on Bayesian optimization is proposed. Experiments are conducted based on the derived early-stage ITSC fault-severity indicator, validating the effectiveness of the proposed fault-diagnosis algorithm.

FTAN: Feature Transform and Alignment Network for cross-domain specific emitter identification

Conventional deep learning-based specific emitter identification (SEI) methods are consistently constrained to domain-invariant assumption, leading to a decrease in recognition accuracy when the feature domain changes. To tackle this issue, we propose a novel unsupervised domain adaptation (UDA) framework named feature transform and alignment network (FTAN) for cross-domain SEI. In FTAN, we first apply a weight-shared network to extract the initial features of signals from all domains. Then, we introduce domain-specific modules to individually learn domain-invariant features, which can minimize the distribution discrepancies of source and target signals. Finally, the aligned domain-invariant features are utilized for identification. We evaluate the performance of FTAN on the various signal datasets. The experimental results demonstrate that FTAN significantly mitigates identification performance degradation in cross-domain scenarios and outperforms other state-of-the-art methods.

Weak ultrasonic guided wave signal recognition based on one-dimensional convolutional neural network denoising autoencoder and its application to small defect detection in pipelines

Pipeline failures are often caused by the expansion of small defects. Structural damage to pipelines can lead to major safety accidents. When ultrasonic guided wave (UGW) technology is used for pipeline failure detection, the echoes produced by small defects manifest as weak UGW signals amidst significant noise. The low amplitude of these signals or complete drowning by noise makes them difficult to recognize. This study innovatively introduces a one-dimensional convolutional neural network denoising autoencoder (1DCNN-based DAE) for noise reduction in UGW signals using deep learning. To improve the conventional DAE, the model incorporated the Parametric Rectified Linear Unit (PReLU) activation function and a CNN for enhanced feature extraction, resulting in the proposed 1DCNN-based DAE. The model is trained on an extensive dataset of mixed signals with strong noise and their corresponding clean signals, enabling autonomous denoising in an unsupervised manner. Additionally, this paper proposes the application of the window-shifted power spectrum method for analyzing the denoised signals to identify and locate pipeline defects. The method involves traversing the signal with a window to intercept fragments, calculating their power, and plotting the power spectrum curve. Defects are then located based on the peak positions of this curve. Numerical simulation and experimental signals were used to validate the proposed method. Simulation results showed that the proposed 1DCNN-based DAE effectively improved the signal-to-noise ratio (SNR) of UGW mixed signals from −9 dB to 21.63 dB, representing an improvement of up to 30.63 dB. Experimental results demonstrated that the method accurately detected weak UGW signals from small defective pipes with a 2 % cross-section loss rate, achieving over 90 % recognition confidence and less than 1.5 % axial positioning error rate. In summary, the proposed 1DCNN-based DAE can effectively improve the SNR of the signal, reduce the noise in the UGW detection signal, and improve the sensitivity of defect identification; the window-shifted power spectrum method has a advantage in the accurate localization of defects.

The Development of Machine Learning Models for Radial Compressor Monitoring With Instability Detection

Abstract Modern radial compressors are designed to operate with high performance and a wide range of applications while minimizing their environmental impact. It is necessary to study the machine near the stability limit and gain a comprehensive understanding of the fluid dynamic mechanisms that trigger the instability in the system to extend the operating range at high efficiency. Machine learning aids in developing pattern identification models for detecting compressor instability. In a prior study, a two-stage radial compressor for refrigerant gas underwent extensive simulation, capturing unsteady RANS conditions and generating a substantial dataset of pressure signals from multiple probes. Selected signals, coupled with detailed computational fluid dynamics (CFD) post-processing, revealed fluid dynamic structures near surge conditions. This study utilizes all pressure signals to assess the stage's operational state (stable, transient, or unstable/surge). An additional goal is to develop an algorithm predicting flow patterns in different stage sections with minimal pressure signals at the rotor inlet. This is a preliminary step toward the development of a smart monitoring and diagnostic model for the compressor installed in a plant. The developed model consists of three submodels, trained with CFD results obtained from unsteady Reynolds averaged Navier–Stokes (URANS) simulations. The three submodels are a regression submodule, a classification submodule, and a forecasting algorithm. The regression submodule predicts diffuser static pressure fields, the classification submodule categorizes whether the compressor is in a critical condition or not, and the forecasting algorithm predicts the inducer pressure signals in the future impeller rotations according to the actual operation history. These sub-modules work together to forecast whether the compressor, according to the operation strategy, is going into instability conditions.

Seamless Optimization of Wavelet Parameters for Denoising LFM Radar Signals: An AI-Based Approach

Linear frequency modulation (LFM) signals are pivotal in radar systems, enabling high-resolution measurements and target detection. However, these signals are often degraded by noise, significantly impacting their processing and interpretation. Traditional denoising methods, including wavelet-based techniques, have been extensively used to address this issue, yet they often fall short in terms of optimizing performance due to fixed parameter settings. This paper introduces an innovative approach by combining wavelet denoising with long short-term memory (LSTM) networks specifically tailored for LFM signals in radar systems. By generating a dataset of LFM signals at various signal-to-noise Ratios (SNR) to ensure diversity, we systematically identified the optimal wavelet parameters for each noisy instance. These parameters served as training labels for the proposed LSTM-based architecture, which learned to predict the most effective denoising parameters for a given noisy LFM signal. Our findings reveal a significant enhancement in denoising performance, attributed to the optimized wavelet parameters derived from the LSTM predictions. This advancement not only demonstrates a superior denoising capability but also suggests a substantial improvement in radar signal processing, potentially leading to more accurate and reliable radar detections and measurements. The implications of this paper extend beyond modern radar applications, offering a framework for integrating deep learning techniques with traditional signal processing methods to optimize performance across various noise-dominated domains.

Tactile-GAT: tactile graph attention networks for robot tactile perception classification

As one of the most important senses in human beings, touch can also help robots better perceive and adapt to complex environmental information, improving their autonomous decision-making and execution capabilities. Compared to other perception methods, tactile perception needs to handle multi-channel tactile signals simultaneously, such as pressure, bending, temperature, and humidity. However, directly transferring deep learning algorithms that work well on temporal signals to tactile signal tasks does not effectively utilize the physical spatial connectivity information of tactile sensors. In this paper, we propose a tactile perception framework based on graph attention networks, which incorporates explicit and latent relation graphs. This framework can effectively utilize the structural information between different tactile signal channels. We constructed a tactile glove and collected a dataset of pressure and bending tactile signals during grasping and holding objects, and our method achieved 89.58% accuracy in object tactile signal classification. Compared to existing time-series signal classification algorithms, our graph-based tactile perception algorithm can better utilize and learn sensor spatial information, making it more suitable for processing multi-channel tactile data. Our method can serve as a general strategy to improve a robot’s tactile perception capabilities.