Pattern Recognition Algorithms Research Articles

Signal Processing for Spinal Cord Injury Monitoring with sEMG

Spinal cord injuries (SCIs) significantly impair motor and sensory functions, posing challenges to rehabilitation and patient management. Accurate monitoring of SCIs is crucial for evaluating the extent of injury and guiding therapeutic interventions. This study explores the application of surface electromyography (sEMG) as a non-invasive tool for monitoring muscle activity in individuals with SCIs. By capturing electrical signals generated by muscle fibers, sEMG provides real-time data essential for assessing neuromuscular function. To enhance the effectiveness of sEMG in spinal cord injury monitoring, advanced signal processing techniques are employed. These techniques include noise reduction, which improves signal clarity, feature extraction to identify key patterns in muscle activation, and pattern recognition algorithms to classify muscle responses. The integration of these methods enables more accurate interpretations of sEMG data, facilitating a better understanding of motor capabilities and rehabilitation progress. The findings highlight the potential of using signal processing in conjunction with sEMG to improve patient outcomes, optimize rehabilitation protocols, and provide insights into the underlying mechanisms of spinal cord injuries. This approach not only advances clinical practices but also supports ongoing research aimed at enhancing the quality of life for individuals affected by SCIs.

Novel application of convolutional neural networks for artificial intelligence-enabled modified moving average analysis of P-, R-, and T-wave alternans for detection of risk for atrial and ventricular arrhythmias

BackgroundT-wave alternans (TWA) analysis was shown in >14,000 individuals studied worldwide over the past two decades to be a useful tool to assess risk for cardiovascular mortality and sudden arrhythmic death. TWA analysis by the modified moving average (MMA) method is FDA-cleared and CMS-reimbursed (CAG-00293R2). ObjectiveBecause the MMA technique is inherently suitable for dynamic tracking of alternans levels, it was selected for development of artificial intelligence (AI)-enabled algorithms using convolutional neural networks (CNN) to achieve rapid, efficient, and accurate assessment of P-wave alternans (PWA), R-wave alternans (RWA), and TWA. MethodsThe novel application of CNN algorithms to enhance MMA analysis generated efficient and powerful pattern-recognition algorithms for highly accurate alternans quantification. Algorithm reliability and accuracy were verified using simulated ECGs achieving R2 ≥ 0.99 (p < 0.01) in response to noise inputs and artifacts that emulate real-life conditions. ResultsAccuracy of the new AI-MMA algorithms in TWA analysis (n = 5) was significantly improved over unsupervised, automated MMA output (p = 0.036) and did not differ from conventional MMA analysis with expert overreading (p = 0.21). Accuracy of AI-MMA in PWA analysis (n = 45) was significantly improved over unsupervised, automated MMA output (p < 0.005) and did not differ from conventional MMA analysis with expert overreading (p = 0.89). TWA and PWA by AI-MMA were correlated with conventional MMA output over-read by an expert reader (R2 = 0.7765, R2 = 0.9504, respectively). ConclusionThis novel technique for AI-MMA analysis could be suitable for use in diverse in-hospital and out-of-hospital monitoring systems, including cardiac implantable electronic devices and smartwatches, for tracking atrial and ventricular arrhythmia risk.

A study of volatiles of young citrus fruits from four areas based on GC–MS and flash GC e-nose combined with multivariate algorithms

The present study has successfully established a scientific and precise approach for distinguishing the geographical origins of young citrus fruits (Qingpi) from four primary production regions in China, using gas chromatography-mass spectrometry (GC–MS) and flash gas chromatography electronic nose (flash GC e-nose) to analyze the volatile composition and odor characteristics. Through the application of chemometric analysis, a clear differentiation among Qingpi samples was established using GC–MS. Additionally, the application of flash GC e-nose facilitated the extraction of flavor information, which enabled the discrimination of geographical origins. Several flavor components were identified as significant factors for origin certification. Furthermore, two pattern recognition algorithms were employed to achieve high accuracy in regional identification. The results of this investigation demonstrate that the amalgamation of multivariate chemometrics and algorithms can proficiently discern the sources of those young citrus fruits. The findings of this research can provide a reference for the assessment of quality control in food and other agricultural commodities in the times ahead.

Machine learning-assisted pattern recognition algorithms for estimating ultimate tensile strength in fused deposition modelled polylactic acid specimens

ABSTRACT In this study, we investigate the application of supervised machine learning algorithms for estimating the Ultimate Tensile Strength (UTS) of Polylactic Acid (PLA) specimens fabricated using the Fused Deposition Modeling (FDM) process. 31 PLA specimens were prepared, with Infill Percentage, Layer Height, Print Speed, and Extrusion Temperature serving as input parameters. The primary objective was to assess the accuracy and effectiveness of four distinct supervised classification algorithms, namely Logistic Classification, Gradient Boosting Classification, Decision Tree, and K-Nearest Neighbor, in predicting the UTS of the specimens. The results revealed that while the Decision Tree and K-Nearest Neighbor algorithms achieved an F1 score of 0.71, the KNN algorithm exhibited a higher Area Under the Curve (AUC) score of 0.79, outperforming the other algorithms. The findings offer valuable insights into the potential use of machine learning techniques in improving the performance and accuracy of predictive models in additive manufacturing.

Utilizing a knowledge-based training algorithm and time-domain extraction for pattern recognition in cylindrical features through vibration and sound signals

This study presents a new solution to address challenges encountered in additive manufacturing, specifically in the context of 3D printing, where failures can occur due to complications associated with the nozzle or filament. The proposed solution in this research involves using a time-domain feature extraction method that leverages sound and vibration patterns. By implementing sensors to capture these signals in a controlled and noise-free environment, and then utilizing a Multi-Layer Perceptron (MLP) model trained accurately to predict upcoming signals and vibrations, proactive anticipation of printing outcomes is facilitated, including potential failures. Simulation results obtained using MATLAB for the MLP showcase the effectiveness of this approach, demonstrating remarkably low error rates. Furthermore, through rigorous data validation, the proposed method's ability to accurately identify sound and vibration signals is confirmed. As a result, the likelihood of failures is significantly reduced, thereby preventing defects in the filament. The implications of this solution hold great promise in substantially enhancing the reliability and efficiency of additive manufacturing processes.

Metal-decorated 3D tin oxide nanotubes in a monolithic sensor array chip for room-temperature gas identification

Inadequate limit of detection at room temperature and unsatisfactory selectivity remain challenge for wide applications of the chemiresistive gas sensors. Nanostructured gas sensors ensure the desirable activation energy for the gas adsorption and chemical reactions, favorable for room-temperature parts-per-billion (ppb) level gas detection. In this work, we demonstrated a single-chip integrated gas sensor array (4 ×4 pixels) based on kinds of metal (Pt, Pd, Au, Ag) decorated 3D SnO2 nanotubes for ultrasensitive room-temperature gas-sensing. The nanostructured sensor array has a high surface area-to-volume ratio, enabling room temperature sensing with high detection response toward H2 (minimum 5 ppm), formaldehyde (minimum 50 ppb), toluene (minimum 50 ppb) and NO2 (minimum 100 ppb) with the detection accuracy within 5% range, respectively. In addition, the effect of metals decoration on the gas identification features were systematically conducted, and these specific response features can be mainly attributed to metal activation capability toward the gases, which enables demonstrating differentiable response patterns for the target gases identification employing a pattern recognition algorithm. These results demonstrate that the proposed strategy helps to provide an excellent route for the future low-power-consumption smart sensing devices design and fabrication.

A comparative optimization procedure to evaluate pattern recognition algorithms on hannes prosthesis

Stability and repeatability of Pattern Recognition (PR) myoelectric control for upper limb prosthetic devices remain unresolved challenges in multi-DoFs systems. In this study, we tested several state-of-the-art classifiers to compare their offline performance in different configurations. Parameters such as realization costs, overall encumbrance, and algorithm complexity were considered for the analysis. The results showed that NLR performed comparably to LDA but with fewer EMG sensors. This study demonstrated that sensor numbers can be reduced to a few units for various algorithms, with NLR being the most tolerant due to its non-linearity. In conclusion, NLR can effectively control the multi-DoFs Hannes system in real-time, offering similar performances to other algorithms while reducing the system's complexity and encumbrance as compared to LDA. It also offers improved tolerance to reduced available information and lower implementation costs.

Copy-Move Forgery Detection in Medical Images Using Handcrafted Features

Abstract: In the realm of medical imaging, the authenticity and integrity of images are paramount for accurate diagnosis and treatment planning. Copy-move forgery, a prevalent form of image tampering, poses a significant threat to the reliability of medical images. This research project focuses on the development and implementation of a robust copy-move forgery detection system tailored specifically for medical images. The proposed methodology leverages handcrafted features, extracting distinctive characteristics from the images to detect instances of forgery. Through a meticulous process of feature engineering and selection, the algorithm aims to enhance sensitivity and specificity in identifying manipulated regions within medical images. The study explores the application of advanced image processing techniques and pattern recognition algorithms to achieve a high level of accuracy in forgery detection.

Design of Electronic Voltage Transformer Error Pattern Recognition and Classification Algorithm Based on Data Mining

With the development of smart grids, electronic voltage transformer (EVT) has gradually entered the stage of large-scale applications. Accurately identifying errors in electronic voltage transformers is crucial for the stability of power systems. Strengthening the measurement accuracy of EVT is of great significance for the operation of power systems and measurement and protection devices. However, due to the limitations of traditional verification methods, there are still challenges. To better improve the accuracy of transformer identification, a data-driven method for enhancing transformer error evaluation and prediction was developed. Based on the low accuracy of traditional EVT error verification and the difficulty of monitoring, data mining technology is proposed for EVT error analysis and evaluation. Recursive principal component analysis is used to separate errors from EVT measurement data, and feature statistics are used to monitor its operating status. Then, regression analysis under support vector machines is added to predict errors for active error correction and better evaluation of its status. The evaluation of the transformer monitoring dataset shows that the classification accuracy of error detection of the proposed method exceeds 93%, and the deviation between the predicted error value and the actual error value is less than 0.05%. Compared with methods such as artificial neural networks and ARMA, the average error rate has been reduced by more than 18%. The accuracy and average accuracy of the algorithm proposed in the study exceeded 80%, with values of 96.23% and 85.12%, respectively. The average error of the ratio difference feature of the EVT is only 0.023, and the average error of the angle difference is less than 0.01, which is much smaller than the algorithm used for comparison. The application response time is less than 0.1 s and the evaluation threshold can better identify data anomalies, with high application accuracy. This method can effectively provide real-time evaluation tools for the operational status of electronic voltage transformers and more accurately and proactively identify transformer errors from conventional data. This study provides an important data-driven solution for improving power grid reliability.

Optimizing the performance of machine learning algorithms for the condition assessment of utility timber poles

ABSTRACT This study focuses on evaluating several machine learning algorithms for the condition assessment of utility timber poles. An efficient feature extraction technique combining Hilbert Huang transform (HHT) and wavelet packet transform (WPT) is adopted from the authors’ previous work and implemented to determine damage-sensitive features from vibration data related to five serviceable and eight unserviceable in-situ timber poles. Then, these features are pre-processed using correlation heat map analysis for feature selection. Principal component analysis (PCA) is adopted as the final step of pre-processing for reducing noise interference and enhancing classification accuracy. Afterwards, a feature matrix is formed, which is fed into the selected classifiers for pattern recognition. In addition, the information gain method is also implemented and compared against PCA to examine the effect of feature selection. Finally, selected classifiers are employed using those dominant features, and their performance is evaluated based on six parameters – accuracy, precision, recall, F1-score, confusion matrix and Receiver Operating Characteristic (ROC) curve. It was found that choosing the best feature set related to the health state helps to improve the performance of the pattern recognition algorithms to a great extent.

Improved deep bidirectional recurrent neural network for learning the cross-sensitivity rules of gas sensor array

Cross-sensitivity among chemical gas sensors leads to inaccurate identification of mixed gas. Pattern recognition algorithms are usually applied to improve the recognition accuracy. However, the abilities of current methods to process sequence property of response data are not strong enough. Especially, they cannot deal with bidirectional cross-sensitive issue, leading to recognition errors. Bidirectional Recurrent Neural Network (BRNN) could well learn bidirectional association features between word sequences in natural language processing field, which is similar to the bidirectional interaction of cross-sensitivity. In this study, an improved deep BRNN model was constructed to solve the cross-sensitivity problem of chemical gas sensor array. A chemical gas sensor array with four units was fabricated and response data was thoroughly obtained. Data preprocessing methods, model structure hyperparameters and optimizers were studied. Finally, an improved deep BRNN model was developed with 3 layers and 100 hidden_size, training with Adamax optimizer. A recognition accuracy of 98.93% was achieved, attributing to the model’s excellent learning ability to the bidirectional cross-sensitivity rules among gas sensors. This improved BRNN model provided a novel idea to eliminate cross-sensitivity, exhibiting good potential for recognizing mixed gas analyte accurately.

Online detection method for variable load conditions and anomalous sound of hydro turbines using correlation analysis and PCA-adaptive-K-means

The hydro turbine is the critical component of hydropower stations. Sound analysis provides a versatile non-contact approach for detecting anomalies or faults of hydro turbines. However, sound analysis methods such as Fourier transform and AI-based pattern recognition algorithms have limitations in achieving online detection due to potential fault frequency aliasing, non-obvious fault characteristics, lack of anomalous samples, and irregular occurrence of anomalous sounds. To overcome these challenges, the proposed method initially employs online correlation analysis to distinguish variable load conditions, followed by the detection of anomalous sounds under stable load conditions using principal component analysis (PCA) and adaptive K-means clustering (K-means). In the method, time–frequency analysis and wavelet packet decomposition are incorporated to extract signal features, with PCA employed to highlight anomalous features. Subsequently, feature correlation analysis and online adaptive clustering are used to accurately detect variable load conditions and anomalous sounds. The proposed method does not rely on prior knowledge of anomalous data, making it flexible and applicable to most practical hydro turbines. Additionally, the method aims to establish a robust foundation for fault diagnosis utilizing sound signals in hydro turbines, providing real-time detection of variable load conditions or anomalous sounds every second. To validate the effectiveness of the proposed method, detection experiments are conducted in practical hydropower station.

Fusion Neural Network for Gas Concentration Prediction in Mixed Gas Environments

Due to the inherent complexity and nonlinearity of mixed gas data, existing pattern recognition algorithms utilized in electronic noses often encounter difficulties in accurately predicting gas concentrations. Addressing this issue, we propose a fusion neural network that merges Long Short-Term Memory (LSTM) and Temporal Convolutional Network (TCN), which we denote as the LSTM-TCN fusion model. The LSTM module effectively captures long-term dependencies in time-series data, while the TCN targets local correlations, thereby enhancing the prediction accuracy for complex gas concentrations. Experimental validation was conducted using a mixed gas dataset comprising ethylene and carbon monoxide. When compared with traditional models, including LSTM, TCN, and GRU, the proposed LSTM-TCN model demonstrated superior performance, achieving an R2 value as high as 0.9922. This research holds considerable practical significance and shows promising application prospects, contributing novel insights and methods to the study and application of electronic nose technology.

Damage Identification of Railway Bridges through Temporal Autoregressive Modeling.

The damage identification of railway bridges poses a formidable challenge given the large variability in the environmental and operational conditions that such structures are subjected to along their lifespan. To address this challenge, this paper proposes a novel damage identification approach exploiting continuously extracted time series of autoregressive (AR) coefficients from strain data with moving train loads as highly sensitive damage features. Through a statistical pattern recognition algorithm involving data clustering and quality control charts, the proposed approach offers a set of sensor-level damage indicators with damage detection, quantification, and localization capabilities. The effectiveness of the developed approach is appraised through two case studies, involving a theoretical simply supported beam and a real-world in-operation railway bridge. The latter corresponds to the Mascarat Viaduct, a 20th century historical steel truss railway bridge that remains active in TRAM line 9 in the province of Alicante, Spain. A detailed 3D finite element model (FEM) of the viaduct was defined and experimentally validated. On this basis, an extensive synthetic dataset was constructed accounting for both environmental and operational conditions, as well as a variety of damage scenarios of increasing severity. Overall, the presented results and discussion evidence the superior performance of strain measurements over acceleration, offering great potential for unsupervised damage detection with full damage identification capabilities (detection, quantification, and localization).

Adaptive marine traffic behaviour pattern recognition based on multidimensional dynamic time warping and DBSCAN algorithm

Identifying marine traffic behaviour patterns is important for intelligent maritime traffic management systems. The introduction of Automatic Identification System (AIS), which can record speed over ground (SOG) and course over ground (COG) data, makes these patterns accessible. Based on that, a novel adaptive marine traffic behaviour pattern recognition algorithm is proposed to explore potential traffic behaviour patterns. In the proposed recognition algorithm, multidimensional dynamic time warping is introduced to measure the similarity among trajectories; this process considers both the spatial and motion characteristics of trajectories. However, the computational complexity of multidimensional dynamic time warping is high. To reduce the number of required computations, a data simplification algorithm considering ship behaviours is adopted to compress trajectories and, therefore, to achieve an improved computation ability. The density-based spatial clustering of applications with noise (DBSCAN) algorithm is utilized to classify marine traffic behaviour patterns. Due to the poor adaptation of the DBSCAN algorithm, we design a local mean density to evaluate the trajectory density distribution in the given dataset and then determine the optimal parameters in DBSCAN according to the local mean density gradient to improve the adaptivity of the algorithm. The proposed approach is illustrated by numerical experiments involving U.S. east coastal waters. The experimental results indicate that our approach is more accurate and adaptive in terms of recognizing hidden traffic behaviour patterns than other methods.

The large-area hybrid-optics CLAS12 RICH: First years of data-taking

The CLAS12 deep-inelastic scattering experiment at the upgraded 12 GeV continuous electron beam accelerator facility of Jefferson Lab conjugates luminosity and wide acceptance to study the 3D nucleon structure in the yet poorly explored valence region, and to perform precision measurements in hadron spectroscopy. A large area ring-imaging Cherenkov detector has been designed to achieve the required hadron identification in the momentum range from 3 GeV/c to 8 GeV/c, with the kaon rate about one order of magnitude lower than the rate of pions and protons. The adopted solution comprises aerogel radiator and composite mirrors in a novel hybrid optics design, where either direct or reflected light could be imaged in a high-packed and high-segmented photon detector. The first RICH module was assembled during the second half of 2017 and installed at the beginning of January 2018, in time for the start of the experiment. The second RICH module, planned with the goal to be ready for the beginning of the operation with polarized targets, has been timely built despite the complications caused by the pandemic crisis and successfully installed in June 2022. The detector performance is here discussed with emphasis on the operation and stability during the data-taking, calibration and alignment procedures, reconstruction and pattern recognition algorithms, and particle identification.

Automated Classification of Dyadic Conversation Scenarios using Autonomic Nervous System Responses.

Two people's physiological responses become more similar as those people talk or cooperate, a phenomenon called physiological synchrony. The degree of synchrony correlates with conversation engagement and cooperation quality, and could thus be used to characterize interpersonal interaction. In this study, we used a combination of physiological synchrony metrics and pattern recognition algorithms to automatically classify four different dyadic conversation scenarios: two-sided positive conversation, two-sided negative conversation, and two one-sided scenarios. Heart rate, skin conductance, respiration and peripheral skin temperature were measured from 16 dyads in all four scenarios, and individual as well as synchrony features were extracted from them. A two-stage classifier based on stepwise feature selection and linear discriminant analysis achieved a four-class classification accuracy of 75.0% in leave-dyad-out crossvalidation. Removing synchrony features reduced accuracy to 65.6%, indicating that synchrony is informative. In the future, such classification algorithms may be used to, e.g., provide real-time feedback about conversation mood to participants, with applications in areas such as mental health counseling and education. The approach may also generalize to group scenarios and adjacent areas such as cooperation and competition.

Deep Learning Logging Sedimentary Microfacies via Improved U-Net

Well logging data contain abundant information on stratigraphic sedimentology. Artificial identification is usually strongly subjective and time-consuming. Pattern recognition algorithms like SVM may not adequately capture the depth-related variations in logging curve shape. This paper defines logging sedimentary microfacies as unidirectional 2D image segmentation and builds an improved U-net model to meet the requirements of logging sedimentary microfacies acquaintance. The proposed model contains three characteristics: (1) It removes pooling layers to avoid the loss of spatial features; (2) it utilizes multi-scale convolution blocks for mining multi-scale spatial features in logging data; (3) a one dimensional convolution layer is added to achieve deep single-direction segmentation. In this model, a 2D image composed of several standardized logging curves is used as the network’s input. In addition, we propose an effective data enhancement method and calculate the geometric feature attributes of well logging curves to reduce the complexity of the data characteristics. We tested the model on manually annotated validation datasets. Our method automatically measures fine sedimentary microfacies characteristics, improving the accuracy of sedimentary microfacies identification and achieving the desired result. Additionally, the model was tested on unlabeled actual logging data, which shows the generalizability of this deep learning method on different datasets.

Human Motion Pattern Recognition Based on Nano-sensor and Deep Learning

A human motion pattern recognition algorithm based on Nano-sensor and deep learning is studied to recognize human motion patterns in real time and with high accuracy. First, human motion data are collected by micro electro mechanical system, and the noise in such data is filtered by smoothing filtering method to obtain high-quality motion data. Second, key time-domain features are extracted from high-quality motion data. Finally, after fusing and processing the key time-domain features, it is input into the deep long and short-term memory (LSTM) neural network to build a deep LSTM human motion pattern recognition model and complete human motion pattern recognition. The results show that the proposed algorithm can realize the recognition of various motion patterns with high accuracy of data acquisition, the average recognition accuracy is 94.8%, the average recall reaches 89.7%, and the F1 score of the algorithm are high, and the recognition time consuming is short, which can realize accurate and efficient human motion pattern recognition and provide guarantee for effective monitoring of the target human motion health.

A deep neural network with electronic nose for water stress prediction in Khasi Mandarin Orange plants

Water stress is a significant environmental factor that hampers plant productivity and leads to various physiological and biological changes in plants. These include modifications in stomatal conductance and distribution, alteration of leaf water potential & turgor loss, altered chlorophyll content, and reduced cell expansion and growth. Additionally, water stress induces changes in the emission of volatile organic compounds across different parts of the plants. This study presents the development of an electronic nose (E-nose) system integrated with a deep neural network (DNN) to detect the presence and levels of water stress induced in Khasi Mandarin Orange plants. The proposed approach offers an alternative to conventional analytical methods that demand expensive and complex laboratory facilities. The investigation employs the leaf relative water content (RWC) estimation, a conventional technique, to evaluate water stress induction in the leaves of 20 plants collected over a span of 9 days after stopping irrigation. Supervised pattern recognition algorithms are trained using the results of RWC measurement, categorising leaves into non-stressed or one of four stress levels based on their water content. The dataset used for training and optimising the DNN model consists of 27 940 samples. The performance of the DNN model is compared to traditional machine learning methods, including linear and radial basis function support vector machines, k-nearest neighbours, decision tree, and random forest. From the results, it is seen that the optimised DNN model achieves the highest accuracy of 97.59% in comparison to other methods. Furthermore, the model is validated on an unseen dataset, exhibiting an accuracy of 97.32%. The proposed model holds the potential to enhance agricultural practices by enabling the detection and classification of water stress in crops, thereby aiding in water management improvements and increased productivity.