Features For Classification Research Articles

Visualization and Analysis of Vocal Fold Dynamics in Various Voice Disorders

Objectives/HypothesisThis study aims to compare the glottal dynamics of vocal fold leukoplakia, laryngitis, and papilloma using various visualization techniques, including phonovibrograms (PVG) and glottovibrograms (GVG). It hypothesizes that these techniques can provide understanding of the vibratory dynamics of these voice disorders, which can aid in their objective diagnosis and deriving visual features for computer-assisted classifications. Study DesignThis work employs a comparative analyses of vocal fold disorders using visualization techniques. The study examines phonovibrograms (PVG), derivative of PVG (PVG-1), digital kymograms (DKG) and glottovibrograms (GVG) to analyze the glottal dynamics of vocal fold leukoplakia, laryngitis, and papilloma. MethodsThe study utilizes high-speed video (HSV) endoscopy to capture vocal fold behavior in different pathological conditions. PVG, GVG and DKG techniques are applied to visualize and analyze the vibratory patterns of the vocal folds. PVG offers a comprehensive portrayal of vocal fold motion, while GVG presents the distance between vocal fold contours. The study also examines PVG-1 as a derivative of PVG. ResultsDistinct modifications in vocal fold vibrations across various stages (open, closed, closing, and opening) are documented using PVG, PVG-1, GVG, DKG, and PVG trajectory. Healthy vocal folds exhibit consistent periodic patterns, while disorders like leukoplakia, laryngitis, and papilloma manifest unique features due to structural and functional variations. The methods employed identify both structural irregularities and functional deviations in vocal fold vibrations. ConclusionsComparative analyses using PVG and other visualization techniques aids in understanding the glottal dynamics of voice disorders. While laryngoscopy and stroboscopy are valuable for real-time assessment, offline analysis techniques like PVG and GVG are useful for tracking glottal dynamics during phonation and assessing treatment effectiveness over time. Studies like this will contribute to standardizing the PVG based diagnostic criteria for its clinical utility, and development of classification tools for improved diagnosis and management of vocal fold pathologies.

Sound classification with time-frequency features in forest environment

The study of forest sound classification has drawn more attention recently due to its potential for illegal activities and natural disaster monitoring. Based on the forest sound classification dataset (FSC22), a dataset specific to possible sound existing in the forest, five classification methods are utilized to investigate the relationship between recognition accuracy and the number of sound acoustic features, as well as the number of target classes. The results confirmed that extreme random forest is the best method for forest sound classification, with an accuracy of around 70% when the target class number is above 20. Further, Mel-frequency cepstral coefficients are the critical feature for sound classification, while fuzzy labels in the dataset may reduce the success rate of recognition.

Innovative methods for microplastic characterization and detection: Deep learning supported by photoacoustic imaging and automated pre-processing data

Plastic products' widespread applications and their non-biodegradable nature have resulted in the continuous accumulation of microplastic waste, emerging as a significant component of ecological environmental issues. In the field of microplastic detection, the intricate morphology poses challenges in achieving rapid visual characterization of microplastics. In this study, photoacoustic imaging technology is initially employed to capture high-resolution images of diverse microplastic samples. To address the limited dataset issue, an automated data processing pipeline is designed to obtain sample masks while effectively expanding the dataset size. Additionally, we propose Vqdp2, a generative deep learning model with multiple proxy tasks, for predicting six forms of microplastics data. By simultaneously constraining model parameters through two training modes, outstanding morphological category representations are achieved. The results demonstrate Vqdp2's excellent performance in classification accuracy and feature extraction by leveraging the advantages of multi-task training. This research is expected to be attractive for the detection classification and visual characterization of microplastics.

Joint Variational Inference Network for domain generalization

Domain generalization is a machine learning task that involves training a model on a set of domains with the goal of achieving high performance on unseen domains. While most domain generalization methods focus on extracting domain-invariant features, they may ignore domain-specific information beyond object styles which is label-relevant for classification. In this paper, we first propose a two-step data generation process in which the domain label and observed data are sampled from two distributions sequentially, and accordingly develop an end-to-end Joint Variational Inference Network (JVINet) for domain generalization. JVINet is a framework that admits a variational-based discriminative structure, which the domain-specific latent variable is learned and integrated to enhance the feature for classification. When testing on unseen target domains, the potential beneficial domain information is hence utilized to improve generalization ability. The overall objective is to optimize the variational lower bound of the conditional joint likelihood functions for both class and domain labels. We provide theoretical proof that JVINet can achieve an optimal lower bound using a variational-based discriminative approach. To evaluate the effectiveness of our method, we compare it with state-of-the-art methods on both simulated and real-world datasets.

A Convolutional Neural Network Classifies Beat-to-Beat Arterial Pressure Spectrograms and Wavelet Transforms according to Age, Sex, and Metabolic State: Novel Frequencies for Cardiovascular Risk Appraisal

Hemodynamic homeostasis is under the control of multiple systems. Continuous blood pressure (BP) time series carry information relevant to the pathophysiological status of the cardiovascular system. Beside average BP, different measures of variability provide prognostic value for distinguishing between cardiovascular risk states. The latter are derived from time- and frequency- domain analysis of BP recordings and are limited by their capacity to provide cumulative risk appraisal among different risk groups. The prediabetic state is associated with cardiovascular risk which is often described to be age- and sex-specific. In this study, we sought to determine distinctive features of continuous arterial pressure (AP) time-series for risk appraisal in a pre-established prediabetic rat model. To overcome the shortcomings associated with conventional BP variability parameters, we trained a convolutional neural network (CNN) using spectrograms and scalograms generated from short-term Fourier- and Morlet wavelet- transforms of AP time series, respectively, from male and female prediabetic rats fed a mild hypercaloric diet for 12- or 24- weeks and their corresponding controls. The CNN consisted of 5 convolutional layers, separated by batch normalization, ReLU activation, and max pooling. The outcome of the fifth convolutional layer had 40% of the data dropped out and delivered to a fully-connected dense layer and then to a softmax function. The data was divided into 80% training, 10% validation, and 10% test sets. The model classified spectrograms and scalograms with test accuracy of 90%+ for the different binary and multi-class classification tasks. On backward propagation of model scores, different salient frequencies were identified as critical features for classification of AP time-series. The identified frequencies were much higher than those typically used in conventional power spectral density (PSD) analysis. Using 3-way ANOVA or principal component analysis for comparison and clustering of the different cardiovascular risk groups of age, sex, and metabolic disease, respectively, according to PSD at the identified peaks does not yield suffcient risk appraisal. This indicates that the CNN possibly captures non-linear relations which are undetectable using inherently linear methods.The results indicate sex- and age-specific patterns of cardiovascular deterioration in prediabetes. Use of artificial intelligence models for the identification of abnormal BP fluctuation patterns can provide suffcient risk appraisal needed for early intervention, particularly in at-risk patients lacking clinical signs of overt cardiovascular disease. None. This is the full abstract presented at the American Physiology Summit 2024 meeting and is only available in HTML format. There are no additional versions or additional content available for this abstract. Physiology was not involved in the peer review process.

The thalamic covariance network is associated with cognitive deficits in patients with cerebral small vascular disease.

Abnormalities in the gray matter structure of cerebral small vessel disease (CSVD) have been observed throughout the brain. However, whether cortico-cortical connections exist between regions of gray matter atrophy in patients with CSVD has not been fully elucidated. This question was tested by comparing the gray matter covariance networks in CSVD patients with and without cognitive impairment (CI). We performed multivariate modeling of the gray matter volume measurements of 61 patients with CI (CSVD-CI), 85 patients without CI (CSVD-NC), and 108 healthy controls using source-based morphological analysis (SBM) to obtain gray matter structural covariance networks at the population level. Then, correlations between structural covariance networks and cognitive functions were analyzed in CSVD patients. Finally, a support vector machine (SVM) classifier was used with the gray matter covariance network as a classification feature to identify CI among the CSVD population. The results of the analysis of all the subjects showed that compared with healthy controls, the expression of the thalamic covariance network, cerebellum covariance network, and calcarine cortex covariance network was reduced in patients with CSVD. Moreover, CSVD-CI patients showed a significant reduction in the expression of the thalamic covariance network, encompassing the thalamus and the parahippocampal gyrus, relative to CSVD-NC patients, which persisted after excluding CSVD patients with thalamic lacunes. In patients with CSVD, cognitive functions were positively correlated with measures of the thalamic covariance network. More than 80% of CSVD patients with CI were correctly identified by the SVM classifier. Our findings provide new evidence to explain the distribution state of gray matter reduction in CSVD patients, and the thalamic covariance network is the core region for early gray matter reduction during the development of CSVD disease, which is related to cognitive deficits. Reduced expression of thalamic covariance networks may provide a neuroimaging biomarker for the early identification of cognitive impairment in CSVD patients.

The ZTF Source Classification Project. III. A Catalog of Variable Sources

The classification of variable objects provides insight into a wide variety of astrophysics ranging from stellar interiors to galactic nuclei. The Zwicky Transient Facility (ZTF) provides time-series observations that record the variability of more than a billion sources. The scale of these data necessitates automated approaches to make a thorough analysis. Building on previous work, this paper reports the results of the ZTF Source Classification Project (SCoPe), which trains neural network and XGBoost (XGB) machine-learning (ML) algorithms to perform dichotomous classification of variable ZTF sources using a manually constructed training set containing 170,632 light curves. We find that several classifiers achieve high precision and recall scores, suggesting the reliability of their predictions for 209,991,147 light curves across 77 ZTF fields. We also identify the most important features for XGB classification and compare the performance of the two ML algorithms, finding a pattern of higher precision among XGB classifiers. The resulting classification catalog is available to the public, and the software developed for SCoPe is open source and adaptable to future time-domain surveys.

Mapping Quaking Aspen Using Seasonal Sentinel-1 and Sentinel-2 Composite Imagery across the Southern Rockies, USA

Quaking aspen is an important deciduous tree species across interior western U.S. forests. Existing maps of aspen distribution are based on Landsat imagery and often miss small stands (<0.09 ha or 30 m2), which rapidly regrow when managed or following disturbance. In this study, we present methods for deriving a new regional map of aspen forests using one year of Sentinel-1 (S1) and Sentinel-2 (S2) imagery in Google Earth Engine. Using observed annual phenology of aspen across the Southern Rockies and leveraging the frequent temporal resolution of S1 and S2, ecologically relevant seasonal imagery composites were developed. We derived spectral indices and radar textural features targeting the canopy structure, moisture, and chlorophyll content. Using spatial block cross-validation and Random Forests, we assessed the accuracy of different scenarios and selected the best-performing set of features for classification. Comparisons were then made with existing landcover products across the study region. The resulting map improves on existing products in both accuracy (0.93 average F1-score) and detection of smaller forest patches. These methods enable accurate mapping at spatial and temporal scales relevant to forest management for one of the most widely distributed tree species in North America.

Rice Diseases Identification Method Based on Improved YOLOv7-Tiny

The accurate and rapid identification of rice diseases is crucial for enhancing rice yields. However, this task encounters several challenges: (1) Complex background problem: The rice background in a natural environment is complex, which interferes with rice disease recognition; (2) Disease region irregularity problem: Some rice diseases exhibit irregular shapes, and their target regions are small, making them difficult to detect; (3) Classification and localization problem: Rice disease recognition employs identical features for both classification and localization tasks, thereby affecting the training effect. To address the aforementioned problems, an enhanced rice disease recognition model leveraging the improved YOLOv7-Tiny is proposed. Specifically, in order to reduce the interference of complex background, the YOLOv7-Tiny model’s backbone network has been enhanced by incorporating the Convolutional Block Attention Module (CBAM); subsequently, to address the irregularity issue in the disease region, the RepGhost bottleneck module, which is based on structural reparameterization techniques, has been introduced; Finally, to resolve the classification and localization issue, a lightweight YOLOX decoupled head has been proposed. The experimental results have demonstrated that: (1) The enhanced YOLOv7-Tiny model demonstrated elevated F1 scores and mAP@.5, achieving 0.894 and 0.922, respectively, on the rice pest and disease dataset. These scores exceeded the original YOLOv7-Tiny model’s performance by margins of 3.1 and 2.2 percentage points, respectively. (2) In comparison to the YOLOv3-Tiny, YOLOv4-Tiny, YOLOv5-S, YOLOX-S, and YOLOv7-Tiny models, the enhanced YOLOv7-Tiny model achieved higher F1 scores and mAP@.5. The improved YOLOv7-Tiny model boasts a single image inference time of 26.4 ms, satisfying the requirement for real-time identification of rice diseases and facilitating deployment in embedded devices.

DRIVER DROWSINESS DETECTION AND CREATING SAFETY MEASURES USING OPENCV,PYTHON

Accidents due to driver drowsy driving is significantly increasing in high ratio now days. So we proposed an AI based solution to reduce the driver drossy driving accident by alerting the driver when they fall asleep. This is an brief abstract that explains system working and principles. This paper present a novel, Real-Time Driver Drowsiness Detection System(DDDS) that utilizes a multi-model approach for fatigue assessment. The system incorporates computer vision and machine learning techniques to analyze facial features, eye movements, and physical signal(if applicable) captured through a webcam processing techniques enhance data processing techniques enhanced data quality for better analysis.Facial landmarks are identified to locate the eyes, and features like Eye Aspect Ratio(EAR) are used to detect signs of fatigue. The system employs a multimodal approach, utilizing computer vision techniques to analyze facial features and eye movements captured through a webcam. Specifically, the DDDS leverages the Haar cascade algorithm for efficient facial landmark detection and the Histogram of Oriented Gradients (HOG) algorithm for feature extraction from facial regions of interest. This combination allows for accurate localization and analysis of key features like eyes. Additionally, the system explores the potential of physiological data (e.g., heart rate variability) for a more comprehensive analysis (if applicable). Machine learning techniques (if applicable) can be employed on the combined features, including those extracted using Haar and HOG, for enhanced fatigue classification. Upon detecting drowsiness, the DDDS triggers customizable alerts (visual, audio, or haptic) to warn the driver. The system additionally explores the potential of physiological data (e.g., heart rate variability) for fatigue detection, offering a more comprehensive approach (if applicable).A machine learning model can be employed on the combined features for robust fatigue classification. Upon detection of drowsiness,the DDDS triggers customizable alerts ( visual, audio, or haptic) to warn the driver. This system addresses limitations of existing solution by: 1.Real-Time processing:Ensuring immediate detection and response to fatigue signs. 2.Multi feature analysis: Combining facial features, eye movements,and physiological data for robust detection. 3.Customizable alerts: Providing drivers control over alert type for a personalized experience. The evaluation methodology employs a dataset of [data type e.g.., image, videos] to access the system’s performance.Metrics like accuracy,precision,recall,and F1-score will be used to evaluate the effectiveness of the DDDS in detecting drowsy driving. This research aims to contribute to safer roads by developing a comprehensive and adaptable Driver Drowsiness Detection System. II.KEYWORDS: Drowsy Driving, Driver Monitoring,Computer Vision, Machine Learning,Eye Tracking, Fatigue Detection,Physiological Signs(if applicable)

Optimizing latent graph representations of surgical scenes for unseen domain generalization.

Advances in deep learning have resulted in effective models for surgical video analysis; however, these models often fail to generalize across medical centers due to domain shift caused by variations in surgical workflow, camera setups, and patient demographics. Recently, object-centric learning has emerged as a promising approach for improved surgical scene understanding, capturing and disentangling visual and semantic properties of surgical tools and anatomy to improve downstream task performance. In this work, we conduct a multicentric performance benchmark of object-centric approaches, focusing on critical view of safety assessment in laparoscopic cholecystectomy, then propose an improved approach for unseen domain generalization. We evaluate four object-centric approaches for domain generalization, establishing baseline performance. Next, leveraging the disentangled nature of object-centric representations, we dissect one of these methods through a series of ablations (e.g., ignoring either visual or semantic features for downstream classification). Finally, based on the results of these ablations, we develop an optimized method specifically tailored for domain generalization, LG-DG, that includes a novel disentanglement loss function. Our optimized approach, LG-DG, achieves an improvement of 9.28% over the best baseline approach. More broadly, we show that object-centric approaches are highly effective for domain generalization thanks to their modular approach to representation learning. We investigate the use of object-centric methods for unseen domain generalization, identify method-agnostic factors critical for performance, and present an optimized approach that substantially outperforms existing methods.

Real Time Object Detection System Using Yolov

Improving object detection models in computer vision is a crucial area of focus. Object detection represents a more sophisticated approach to image classification, where a neural network identifies objects within an image and delineates them using bounding boxes. YOLO (You Only Look Once) revolutionized object detection by introducing an end-to-end neural network architecture that simultaneously predicts bounding boxes and class probabilities. Unlike earlier methods that adapted classifiers for detection, YOLO's approach streamlines the process by integrating classification and localization tasks, accurately delineating objects with bounding boxes. The project mainly focuses on augmenting YOLOv4, a state-of-the-art object detection architecture, to address the limitations like small object detection, adaptive lighting handling, occlusion challenges, detecting objects at different angles, and specialized object features. Additionally, we aim to improve the model's ability to recognize specialized object features for superior classification. Furthermore, by incorporating functionalities like license plate recognition using Tesseract OCR, object counting (total and class-wise), and detection cropping, this project extends the model's applicability to real-world scenarios. These functionalities make the model valuable in areas like autonomous vehicles, traffic monitoring, and industrial automation. The augmented model's improvements aim to overcome drawbacks identified in existing research, enhancing its efficacy in various scenarios.

Expert evaluation of dual-use technical products

Subject. Progressive and accelerating economic growth of the Russian military-industrial complex needs additional extra-budgetary revenues, which can be obtained using the creation and market sale of dual-use technical products used in the military and civilian spheres. Objectives. The great number of dual technical products, and the complexity of comprehensive assessment of the possibility of their commercialization necessitates creation of special system-wide information and analytical tools. Methods. For objective economic assessment of this type of products, we developed a conceptual method based on their preliminary clustering (grouping) and systematic classification, taking into account the conditions and possibilities of commercial distribution. Results. We identified the main features of clustering and classification of dual products, i.e. the rate of obtaining additional financial profit and economic return from commercial sales, the amount of necessary investments, risky situations and uncertainties. We developed a method to evaluate their commercial and technical potential. The expert assessment of this type of products helps specify and clarify the possibilities and the most effective directions of their commercialization already at the preliminary stage of the systems analysis, identify additional methods for their further and more accurate analysis. The developed expert assessment method can be used in the military-industrial complex, and by high-tech enterprises producing civilian products. Conclusions. The proposed expert assessment of dual-use technical products will enable to develop a system regulation for performing a pre-investment assessment of future sales of technical products, increase the productivity and profitability of defense enterprises.

Definition, acronyms, nomenclature, and classification of angioedema (DANCE): AAAAI, ACAAI, ACARE, and APAAACI DANCE consensus

BackgroundAngioedema (AE) manifests with intermittent, localized, self-limiting swelling of the subcutaneous and/or submucosal tissue. AE is heterogeneous, can be hereditary or acquired, occurs only once or be recurrent, with or without wheals, due to mast cell mediators, bradykinin or other mechanisms. Currently, different taxonomic systems are used, making it difficult to compare the results of studies, develop multicenter collaboration, and harmonize treatments of AE patients. ObjectiveTo develop a consensus on the definition, acronyms, nomenclature, and classification of angioedema (DANCE). MethodsThe initiative involved 91 experts from 35 countries and was endorsed by 53 scientific, medical societies, and patient organizations. A consensus was reached by online discussion and voting using the Delphi process over a period of 16 months (June 2021 to November 2022). ResultsThe DANCE initiative resulted in an international consensus on the definition, classification and terminology of AE. The new consensus classification features five types and endotypes of AE and a harmonized vocabulary of abbreviations and acronyms. ConclusionThe DANCE classification complements current clinical guidelines and expert consensus recommendations on the diagnostic workup and treatment of AE. DANCE does not replace current clinical guidelines and expert consensus algorithms and should not be misconstrued in a way that affects reimbursement of medicines prescribed by a physician using sound clinical judgment. We anticipate that the new AE taxonomy and nomenclature will harmonize and facilitate AE research and clinical studies, thereby improving patient care.

Classification feature selection and dimensionality reduction based on logical binary sine-cosine function arithmetic optimization algorithm

Arithmetic optimization algorithm (AOA) is a meta-heuristic algorithm inspired by mathematical operations. AOA has been diffusely used for optimization issues on continuous domains, but few scholars have studied discrete optimization problems. In this paper, we proposed Binary AOA (BAOA) based on two strategies to handle the feature selection problem. The first strategy used S-shaped and V-shaped shift functions to map continuous variables to discrete variables. The second strategy was to combine four logical operations (AND, OR, XOR, XNOR) on the basis of the transfer function, and constructed a parameter model based on the sine and cosine function. An enhanced logic binary sine–cosine function arithmetic optimization algorithm (LBSCAOA) was proposed to realize the position update of variables. Its purpose was to improve the algorithm's global search capabilities and local exploitation capabilities. In the simulation experiments, 20 datasets were selected to testify the capability of the proposed algorithm. Since KNN had the advantages of easy understanding and low training time complexity, this classifier was selected for evaluation. The performance of the improved algorithm was comprehensively evaluated by comparing the average classification accuracy, the average number of selected features, the average fitness value and the average running time. Simulation results showed LBSCAOA with V-Shaped “V4” stood out among many improved algorithms. On the other hand, LBSCAOA with V-Shaped “V4” was used as a representative to compare with other typical feature selection algorithms to verify its competitivenes.

Transfer Learning-Based Hyperspectral Image Classification Using Residual Dense Connection Networks.

The extraction of effective classification features from high-dimensional hyperspectral images, impeded by the scarcity of labeled samples and uneven sample distribution, represents a formidable challenge within hyperspectral image classification. Traditional few-shot learning methods confront the dual dilemma of limited annotated samples and the necessity for deeper, more effective features from complex hyperspectral data, often resulting in suboptimal outcomes. The prohibitive cost of sample annotation further exacerbates the challenge, making it difficult to rely on a scant number of annotated samples for effective feature extraction. Prevailing high-accuracy algorithms require abundant annotated samples and falter in deriving deep, discriminative features from limited data, compromising classification performance for complex substances. This paper advocates for an integration of advanced spectral-spatial feature extraction with meta-transfer learning to address the classification of hyperspectral signals amidst insufficient labeled samples. Initially trained on a source domain dataset with ample labels, the model undergoes transference to a target domain with minimal samples, utilizing dense connection blocks and tree-dimensional convolutional residual connections to enhance feature extraction and maximize spatial and spectral information retrieval. This approach, validated on three diverse hyperspectral datasets-IP, UP, and Salinas-significantly surpasses existing classification algorithms and small-sample techniques in accuracy, demonstrating its applicability to high-dimensional signal classification under label constraints.

Detection of Emerging Stress in Trees Using Hyperspectral Indices as Classification Features

This research presents a classification methodology for the detection of new or emerging stress in trees using indices derived from hyperspectral data and tests whether existing hyperspectral indices are effective when used as the classification features for this problem. We tested six existing indices—Water Band Index (WBI), Gitelson–Merzlyak B Index (GMb), Normalized Phaeophytization Index (NPQI), Combined Carotenoid/Chlorophyll Ratio Index (CCRI), Photochemical Reflectance Index (PRI), and Red-Edge Chlorophyll Index (CIre)—along with a seventh Test Index—generated as a composite of PRI and Cire—as classification features. Analysis was conducted using data collected from trees with and without emerald ash borer (EAB) infestation to develop a methodology that could be adapted to measure emerging stress from other pathogens or invasive pests in other tree species. Previous work has focused specifically on the identification of damage or stress symptoms caused by a specific pathogen. In this study, we adapted that work to develop a system of classification that can be applied to the identification of stress symptoms from a range of sources, measurable in trees based on spectral response and, in some cases, detectable prior to the onset of visible symptoms that can be measured through human observation. Our data indicate that existing indices derived from hyperspectral data are effective as classification features when measuring spectral responses indicative of emerging stress in trees.

On the importance of integrating convolution features for Indian medicinal plant species classification using hierarchical machine learning approach

This work proposes a novel hierarchical classification framework designed to categorize hundred Indian medicinal plant species. The innovation lies in introducing a comprehensive feature representation by integrating convolutional features with geometric, texture, shape, and multispectral features for classification tasks. In this study, a two-level hierarchical plant classification model is proposed to address the challenges of inter-class similarity and intra-class variations. The first level classifies 100 medicinal plant species into 11 groups based on visual similarities among the plants. At level two, the specific plant species containing in each group are predicted using Random Forest classifier. The evaluation is performed at two levels to analyze the effectiveness of the proposed model. The performance analysis compares the effectiveness of individual feature types against the composite feature model. Performance is also evaluated based on specific groups that demonstrate high similarity between classes and intra-class variations among the plant species separately. Furthermore, the generality of the model is tested using two self-created datasets-RTL80 and RTP40, requiring more than 300 man-hours to collect. Experimental results demonstrate a promising accuracy of 94.54% on GSL100 leaf dataset and 75.46% on RTL80 and RTP40 real-time datasets reflecting the superiority of the proposed hierarchical model over state-of-the-art methods.

Multi-Stream Isolated Sign Language Recognition Based on Finger Features Derived from Pose Data

This study introduces an innovative multichannel approach that focuses on the features and configurations of fingers in isolated sign language recognition. The foundation of this approach is based on three different types of data, derived from finger pose data obtained using MediaPipe and processed in separate channels. Using these multichannel data, we trained the proposed MultiChannel-MobileNetV2 model to provide a detailed analysis of finger movements. In our study, we first subject the features extracted from all trained models to dimensionality reduction using Principal Component Analysis. Subsequently, we combine these processed features for classification using a Support Vector Machine. Furthermore, our proposed method includes processing body and facial information using MobileNetV2. Our final proposed sign language recognition method has achieved remarkable accuracy rates of 97.15%, 95.13%, 99.78%, and 95.37% on the BosphorusSign22k-general, BosphorusSign22k, LSA64, and GSL datasets, respectively. These results underscore the generalizability and adaptability of the proposed method, proving its competitive edge over existing studies in the literature.

Predicting Fine Dead Fuel Load of Forest Floors Based on Image Euler Numbers

The fine dead fuel load on forest floors is the most critical classification feature in fuel description systems, affecting several parameters in the manifestation of wildland fires. An accurate determination of this fine dead fuel load contributes substantially to effective wildland fire prevention, monitoring, and suppression. This study investigated the viability of incorporating image Euler numbers into predictive models of fine dead fuel load via the taking photos method. Pinus massoniana needles and Quercus fabri broad leaves—typical fuel types in karst areas—served as the research subjects. Accurate field data were collected in the Tianhe Mountain forests, China, while artificial fine dead fuelbeds of differing loads were constructed in the laboratory. Images of the artificial fuelbeds were captured and uniformly digitized according to various conversion thresholds. Thereafter, the Euler numbers were extracted, their relationship with fuel load was analyzed, and this relationship was applied to generate three load-prediction models based on stepwise regression, nonlinear fitting, and random forest algorithms. The Euler number had a significant relationship with both P. massoniana and Q. fabri fuel loads. At low conversion thresholds, the Euler number was negatively correlated with fuel load, whereas a positive correlation was recorded when this threshold exceeded a certain value. The random forest model showed the best prediction performance, with mean relative errors of 9.35% and 14.54% for P. massoniana and Q. fabri, respectively. The nonlinear fitting model displayed the next best performance, while the stepwise regression model exhibited the largest error, which was significantly different from that of the random forest model. This study is the first to propose the use of image features to predict the fine fuel load on a surface. The results are more objective, accurate, and time-saving than current fuel load estimates, benefiting fuel load research and the scientific management of wildland fires.