Euclidean Distance Classifier Research Articles

A sophisticated Drowsiness Detection System via Deep Transfer Learning for real time scenarios

<abstract> <p>Driver drowsiness is one of the leading causes of road accidents resulting in serious physical injuries, fatalities, and substantial economic losses. A sophisticated Driver Drowsiness Detection (DDD) system can alert the driver in case of abnormal behavior and avoid catastrophes. Several studies have already addressed driver drowsiness through behavioral measures and facial features. In this paper, we propose a hybrid real-time DDD system based on the Eyes Closure Ratio and Mouth Opening Ratio using simple camera and deep learning techniques. This system seeks to model the driver's behavior in order to alert him/her in case of drowsiness states to avoid potential accidents. The main contribution of the proposed approach is to build a reliable system able to avoid false detected drowsiness situations and to alert only the real ones. To this end, our research procedure is divided into two processes. The offline process performs a classification module using pretrained Convolutional Neural Networks (CNNs) to detect the drowsiness of the driver. In the online process, we calculate the percentage of the eyes' closure and yawning frequency of the driver online from real-time video using the Chebyshev distance instead of the classic Euclidean distance. The accurate drowsiness state of the driver is evaluated with the aid of the pretrained CNNs based on an ensemble learning paradigm. In order to improve models' performances, we applied data augmentation techniques for the generated dataset. The accuracies achieved are 97 % for the VGG16 model, 96% for VGG19 model and 98% for ResNet50 model. This system can assess the driver's dynamics with a precision rate of 98%.</p> </abstract>

Transfer learning and clustering analysis of epileptic EEG signals on Riemannian manifold

Due to the nonlinearity and non-stationarity of EEG signals, the manifold becomes a new direction to analyze EEG signals. However, the problem of imbalanced epileptic seizure data and limited labeling data results in insufficient training data. The significant difference in feature distribution between cross-subject EEG signals also affects model generalization. A manifold transfer learning method based on divergence alignment (MTLDA) was proposed to overcome the few-shot problem. By constructing an objective function based on domain divergence, the transformation matrix is obtained to reduce the feature distribution difference between source and target domains. Additionally, for less labeled data on the manifold, a Riemannian manifold fuzzy clustering (RMFC) was proposed to realize unsupervised pattern recognition of the sample on the manifold. MTLDA and RMFC use Riemannian distance and Riemannian mean rather than the classical Euclidean distance and mean. Both methods do not require labeling information from subjects. The experiment results on two public datasets and one private dataset show that MTLDA and RMFC are effective and outperform other comparison algorithms.

An Enhanced Prediction Model For Autism Spectrum Disorder

Autism spectrum disorder (ASD) is a behavioural condition that affects the child’s social interaction, communication, and behaviour. The early identification of ASD is critical for the effective and timely therapies. This study presents an enhanced prediction model for Autism Spectrum Disorder (ASD). It is based on facial features extracted from the face image. Mallat’s multi-resolution algorithm is employed in this work for extracting facial features. Two distance based classifiers such as Euclidean Distance Classifier (EDC) and Absolute Distance Classifier (ADC) are employed for the ASD prediction. The proposed ASD prediction system is evaluated on face images of autistic and non-autistic children. The database is obtained from the Kaggle data repository. A total of 2940 facial images (1470 autistic and 1470 non-autistic) are employed for performance analysis. Experimental results show that the proposed ASD prediction system provides promising results with an accuracy of 97.01% by EDC and 96.87% by ADC classifiers. Keywords Autism spectrum disorder, Computer aided diagnosis, Machine learning, Image based diagnosis, Prediction system.

Open-Set Signal Recognition Based on Transformer and Wasserstein Distance

Open-set signal recognition provides a new approach for verifying the robustness of models by introducing novel unknown signal classes into the model testing and breaking the conventional closed-set assumption, which has become very popular in real-world scenarios. In the present work, we propose an efficient open-set signal recognition algorithm, which contains three key sub-modules: the signal representation sub-module based on a vision transformer (ViT) structure, a set distance metric sub-module based on Wasserstein distance, and a class space compression sub-module based on reciprocal point separation and central loss. In this algorithm, the representing features of signals are established based on transformer-based neural networks, i.e., ViT, in order to extract global information about time series-related data. The employed reciprocal point is used in modeling the potential unknown space without using the corresponding samples, while the distance metric between different class spaces is mathematically modeled in terms of the Wasserstein distance instead of the classical Euclidean distance. Numerical experiments on different open-set signal recognition tasks show that the proposed algorithm can significantly improve the recognition efficiency in both known and unknown categories.

Identifikasi Kematangan Buah Menggunakan Metode Gray Level Co-occurence Matrix pada Citra Digital

In the current era of information technology, the use of images is widely applied in various aspects of life. Some of the uses of image processing include the fields of military, medicine, education, agriculture and so on. One example of the use of image processing that will be discussed in this study is in the agricultural sector. Farmers can take advantage of technology in selecting fruit with the appropriate maturity level. In terms of selecting fruit based on the level of maturity, some fruit farmers still use the conventional method or with the human sense of sight, namely the eye. Therefore, this study was conducted for preliminary research that can change the conventional method into system that uses technology that makes a computerized way of identifying fruit maturity levels. The system that will be used to identify fruit maturity uses the Gray Level Co-occurance Matrix method on digital images and Euclidean Distance as a classification method. This application is built using MatLab 2019a programming. The test results show that the Gray Level Co-occurance Matrix and Euclidean Distance methods can be used to identify the ripeness of guava, oranges, bananas, papayas and mangoes into three categories, namely raw, unripe and ripe. The Gray Level Co-occurance Matrix and Euclidean Distance classification methods succeeded in identifying fruit maturity with an overall success rate of 87%.

Population Classification Model of Liaoning Province Based on Cluster Analysis

Aiming at the problem of urban shrinkage in Liaoning Province, this paper established a population classification model by systematic clustering method. Based on two indicators of population contraction and GDP contraction, we defined the shrinkage rate, and classified 30 cities in Liaoning Province according to the shrinkage rate. Firstly, the weights of population contraction and GDP contraction were calculated by using the analytic hierarchy process. Secondly, the average annual growth rate of the two is weighted, and the shrinkage rate is defined as the weighted value. Then, based on the systematic clustering method, the population classification model was established by using the classical Euclidean distance, and the 30 cities were classified by SPSS software. The results of this paper on shrinking cities have important reference value for examining the future development trend of a city.

Content based Image Retrieval Using Clustering Method

Content-based image retrieval (CBIR) is the deployment of computer vision methods to the information retrieval challenge, that is, the subject of seeking out digital images in vast databases. Techniques based on automated feature extraction methods for obtaining similar images from image databases are under the purview of CBIR. Traditional content based image retrieval (CBIR) systems extract a single feature at a time and use it to categorize and group images in response to a query. To bridge the gap between high-level concepts and low-level features, our innovative method integrates many feature extraction algorithms. In color-based retrieval, we use quadratic distance formulas to calculate the HSV affinity matrix for photos in the query and the database. Wavelet decomposition at six stages is used in texture-based retrieval. Finding the similarity measures between the query image and the images in the database is done with the help of the Euclidean distance classifier. The integrated method used to decrease the file sizes of the retrieved photographs keeps the user from having to pay as much attention to the process.

Online automatic monitoring of abnormal vibration of stay cables based on acceleration data from structural health monitoring

Online automatic monitoring of abnormal vibrations, such as vortex-induced vibration (VIV) and high amplitude vibration (HAV), of stay-cables is important for bridge maintenance. However, the existing methods either require manual intervention or yield unreasonable identification results and are not suitable for online automatic monitoring. This study proposes a fully automatic method to identify the abnormal vibrations of stay-cables based on acceleration data. Two quantified feature indices are extracted from frequency domain and complex domain to characterize the VIV. A feature vector is established to distinguish the VIV from ambient vibration, and Minimum Euclidean Distance Classifier is utilized to automate the online identification process. The HAV is automatically identified based on root mean square of acceleration. The proposed method avoids any manual intervention and can yield correct and convincible identification results. Case study of a cable-stayed bridge finally validates the feasibility and novelty of the proposed method.

STUDY OF PHENOTYPIC VARIABILITY USING THE VARIETAL DIVERSITY OF CULTIVATED FORMS OF NAKED AND HULLED OATS IN THE INTERCROPPING SYSTEM

The research carried out at the Gene Bank in Suceava, Northern Romania, aimed to highlight the phenotypic variability of the germplasm of Avena spp. For this purpose, the morpho-productive traits and resistance to diseases, pests and weeds were analysed. Productivity, diseases and pests, days to heading and days to maturity descriptors of 46 Avena spp. genotypes (naked and hulled oat) with different biological statuses (36 local populations, 10 cultivars), were evaluated by testing in intercropping experiments with small grain cereals and grain legumes. The unidirectional ANOVA analysis generated values that allowed the elaboration of a hierarchy of heterogeneity in the hulled local populations, for some of the analysed characters (one thousand seeds weight/genotype, degree of attack by Puccinia coronata and Oulema melanopa and days to maturity) and these were less in naked forms. There was a high competition of Avena species, regarding weeds in small grain cereal variants and potentially beneficial for nitrogen symbiotic fixation by increasing naked and hulled local oat population productivity in intercropping with fava bean. The Euclidean distance classification of the oat genotypes investigated in the dendrogram distinguished the generated groups, indicating the maximum distance in cluster IV. With high heterogeneity of productivity traits, better resistance to disease and pests, and shorter maturity, members of this cluster could be used to develop genetic mixture programs.

A Customer Clustering Algorithm for Power Logistics Distribution Network Structure and Distribution Volume Constraints

Customer clustering technology for distribution process is widely used in location selection, distribution route optimization and vehicle scheduling optimization of power logistics distribution center. Aiming at the problem of customer clustering with unknown distribution center location, this paper proposes a clustering algorithm considering distribution network structure and distribution volume constraint, which makes up for the defect that the classical Euclidean distance does not consider the distribution road information. This paper proposes a logistics distribution customer clustering algorithm, which improves CLARANS algorithm to make the clustering results meet the constraints of customer distribution volume. By using the single vehicle load rate, the sufficient conditions for logistics distribution customer clustering to be solvable under the condition of considering the ubiquitous and constraints are given, which effectively solves the problem of logistics distribution customer clustering with sum constraints. The results state clearly that the clustering algorithm can effectively deal with large-scale spatial data sets, and the clustering process is not affected by isolated customers, The clustering results can be effectively applied to the distribution center location, distribution cost optimization, distribution route optimization and distribution area division of vehicle scheduling optimization.

A discrete adjoint method for pressure-based algorithms

A discrete adjoint method implemented in a coupled pressure-based RANS solver is presented in this paper. The adjoint equations are solved using an adjoint fixed point iteration that inherits the convergence properties of the primal solver. Automatic differentiation is used extensively for the construction of the adjoint fixed point iteration. The concept of Krylov subspace methods was adopted to stabilize the solution procedure. A common linearization technique in collocated pressure-based algorithms is the introduction of a mass flux variable on the cell faces which is kept constant during the inner iterations. This variable is treated as an independent adjoint variable in related publications. We propose a new method that allows to treat the mass fluxes implicitly in order to take full advantage of the preconditioner of the primal solver. The adjoint solver is general and is not restricted by the commonly used frozen turbulence approximation. It can deal with any turbulence model that is supported by the flow solver as well as any boundary condition. This includes mixing planes and mesh interfaces needed for multi stage turbo machinery simulations. Furthermore, there is no restriction on the choice of objective function. The sensitivities of the adjoint solver have been validated with sensitivities obtained with finite differences.An entirely surface based interpolation method based on radial basis functions (RBF) was developed to deform the surface mesh. We propose the use of discrete geodesics instead of the classical Euclidean distance as the distance measure for the RBF interpolation. As an alternative, a direct deformation method with adjoint consistent smoothing is also described and used in the presented optimization cases.The developed adjoint solver and deformation routines were used to optimize a turbulent bend with different Reynolds numbers as well as the the rotor blade of an axial turbine.

Cloud model-clustering analysis based evaluation for ventilation system of underground metal mine in alpine region

Ventilation system is significant in underground metal mine of alpine region. Reasonable evaluation of ventilation effectiveness will lead to a practical improvement for the maintenance and management of ventilation system. However, it is difficult to make an effective evaluation of ventilation system due to the lack of classification criteria with respect to underground metal mine in alpine region. This paper proposes a novel evaluation method called the cloud model-clustering analysis (CMCA). Cloud model (CM) is utilized to process collected data of ventilation system, and they are converted into cloud descriptors by CM. Cloud similarity (CS) based Euclidean distance (ED) is proposed to make clustering analysis of assessed samples. Then the classification of assessed samples will be identified by clustering analysis results. A case study is developed based on CMCA. Evaluation results show that ventilation effectiveness can be well classified. Moreover, CM is used alone to make comparison of evaluation results obtained by CMCA. Then the availability and validity of CMCA is verified. Meanwhile, difference of CS based ED and classical ED is analyzed. Two new clustering analysis methods are introduced to make comparison with CMCA. Then the ability of proposed CMCA to meet evaluation requirements of ventilation system is verified.

Clothing Style Recognition Method Based on Digital Image Processing

The use of digital image processing technology to identify clothing styles from clothing images has broad application potential in clothing consumption analysis, auxiliary clothing design and identity recognition. In the current clothing style recognition field for feature extraction and classification of clothing contours, the main methods are extreme learning machine classification based on wavelet Fourier descriptor, and Euclidean distance classification based on fusion features. However, these methods also have some shortcomings. Currently, there is still no effective method to extract features and classify clothing contours. In order to solve this problem, this paper is based on the SVM classification method of Fourier contour description features. The contour curvature characteristic point in the figure can be described as the bending characteristic of the clothing contour curve, and its calculation is simple and intuitive, which improves the judgment effect of the similarity performance of the Hausdorff distance. This paper constructs a multi-resolution framework of the new image, uses edge detection algorithms to obtain the edges of each level of resolution image, and expands them into narrow edge bands, and combines the point distribution model with the edge narrow bands obtained by edge detection to perform multi-resolution image ASM search for. This paper proposes an image fusion algorithm based on algebraic multigrid and adaptive block. Because the coarse grid extracted by the algebraic multi-grid method can extract the detailed information of the image to a certain extent, the original image can be reconstructed from the grid data. Experimental research shows that from the overall recognition rate of various clothing styles, the Fourier descriptor has a high recognition rate for each clothing, and the stability of multiple experiments is also good.

Automatic Identity Verification Using Face Images.(Dept.E)

This paper presents two types of identity verification systems based on face images. In the first system principle component analysis (PCA) is used for dimensionality reduction and feature extraction whereas in the second system wavelet transform is used for dimensionality reduction and (PCA) for feature extraction. Two types of classifiers, the Euclidean distance and Angle distance, are used with both systems. Comparative study by computer simulation showed that the addition of wavelet transform before (PCA) provides better performance in the case of Euclidean distance classifier. Moreover, the results obtained by (PCA) are in agreement with those reported in literatures.

Wavelet-based features selected with Paraconsistent Feature Engineering successfully classify events in low-voltage grids

Energy quality, either in centralized or distributed generation, is directly affected by events in electrical lines. Consequently, the precise identification of those issues is of paramount importance, where the features extracted from domestic or industrial low-voltage sources should be able to properly represent the events for a subsequent classification. Nevertheless, current algorithms for event diagnosis suffer from a number of drawbacks such as the lack of real data to model the problem, since the majority of strategies is supported by simulated signals, and the uncertainty on the best features to conveniently address the occurrences. Thus, our contribution in this paper is twofold: we describe our own database, which is freely available under request, and innovatively apply Paraconsistent Feature Engineering (PFE) to analyze and select favorite wavelet-based features to classify events in low-voltage grids. Lastly, an example application where a set of features was capable of distinguishing specific events from normal signals with a value of accuracy of 96% using just an Euclidean distance classifier is shown, reassuring the efficacy of the proposed approach. Notably, the association of wavelets with PFE to handle energy quality issues had never been reported in literature.

P.141 Topological data analysis for genetic-driven stratification of patients with major depressive disorder

The clustering of hyperspectral images is a challenging task because of the high dimensionality of the data. Sparse subspace clustering (SSC) algorithm is one of the popularly used clustering algorithm for high dimensionality data. However, SSC has not fully used the spectral and spatial information during similarity matrix construction based on single sparse representation coefficient for hyperspectral Imagery (HSI) clustering. In this paper, two novel similarity matrix construction methods named as Cosine-Euclidean similarity matrix (abbreviated as CE) and Cosine-Euclidean dynamic weighting similarity matrix (abbreviated as CEDW) are proposed for HSI clustering. They can combine the high spectral information and rich spatial information. Firstly, CE utilizes the cosine similarity of spectral information based on overall sparse representation vectors and classical Euclidean distance of spatial information to construct a novel similarity matrix. Secondly, inheriting CE merits, dynamic weighting adjustment method is introduced to CEDW for some external influence factors to the HSI information. Several experiments on HSI demonstrated that the proposed algorithms are effective for HSI clustering.

LEVERAGING THE INTERNET OF THINGS BASED ON THE ANALYTICAL STUDY OF THE SKELETON JOINT POSITION IN RECOGNISING THE QUALITY OF HUMAN ACTIVITY

To give consequently dissecting and identifying human exercises to offer better help in the medical services area, security reason and so on Strategy: We have utilized UTKinect-Action 3D dataset containing the Position of 20 body joint caught by Kinect sensor. We chose two arrangement of joints J1 and J2; after that, we have shaped a few principles for movement grouping then we have applied SVM classifier, KNN classifier utilizing Euclidean distance and KNN classifier using Minkowski distance for action order. At the point when we have been used joint set J1 we got 97.8% exactness with SVM classifier, 98.8% precision with KNN classifier utilizing Euclidean distance, and 98.9% exactness with KNN classifier utilizing Minkowski distance and for joint set J2 we got 97.7% exactness with SVM classifier, 98.6% exactness with KNN classifier using Euclidean distance, and 98.7% exactness with KNN classifier utilizing Minkowski distance. We have arranged four exercises hand waving, standing, sitting and picking. In future, more exercises can likewise be remembered for this examination. IoT alongside this action acknowledgement strategy can be utilized to lessen overheads

A fuzzy machine learning approach for identification of paddy stubble burnt fields

Stubble burning is the most common practice in Northern region of India and is considered as the main reason for pollution in nearby NCR (National Capital Region) New Delhi, India. Therefore, in this research work, paddy stubble burnt fields were identified by using fuzzy based machine learning algorithms for a test site in Patiala, Punjab, India. An Euclidean Distance (ED) and Gaussian Kernel based Modified Possibilistic c-Means (MPCM) algorithms were used for this purpose. The main objective of this research work was to test the effectiveness of machine learning approach for identification of paddy stubble burnt filed and also to find out the better classifier amongst Euclidean Distance (ED) and Gaussian Kernel MPCM classifier. The Sentinel 2A/2B cloud free temporal data having a high frequency of repetitiveness in the interval of 5-days were used. To reduce spectral dimensionality of data Class Based Sensor Independent Indices (CBSII) database were generated. The burning was mainly observed from the last week of October to the first week of November and has been visualized using variation of CBSII-NDVI temporal plot. Normalized Burnt Ratio (NBR) was computed and compared with CBSII-NDVI outputs for both the ED and Kernel based MPCM classifiers. Delta Normalized Burnt Ratio (ΔNBR) was computed for cross-validation of burnt patches and is found to be in between 0.45 and 0.66. Thus, it can be concluded that kernel based MPCM classifiers are more efficient in terms of accuracy and robustness to noise in comparison to distance based MPCM classifiers, for the mapping of paddy burnt fields.

Single Channel Source Separation with ICA-Based Time-Frequency Decomposition.

This paper relates to the separation of single channel source signals from a single mixed signal by means of independent component analysis (ICA). The proposed idea lies in a time-frequency representation of the mixed signal and the use of ICA on spectral rows corresponding to different time intervals. In our approach, in order to reconstruct true sources, we proposed a novelty idea of grouping statistically independent time-frequency domain (TFD) components of the mixed signal obtained by ICA. The TFD components are grouped by hierarchical clustering and k-mean partitional clustering. The distance between TFD components is measured with the classical Euclidean distance and the distance of Gaussian distribution introduced by as. In addition, the TFD components are grouped by minimizing the negentropy of reconstructed constituent signals. The proposed method was used to separate source signals from single audio mixes of two- and three-component signals. The separation was performed using algorithms written by the authors in Matlab. The quality of obtained separation results was evaluated by perceptual tests. The tests showed that the automated separation requires qualitative information about time-frequency characteristics of constituent signals. The best separation results were obtained with the use of the distance of Gaussian distribution, a distance measure based on the knowledge of the statistical nature of spectra of original constituent signals of the mixed signal.

Instance-based learning of marker proteins of carcinoma cells for cell death/ survival

ABSTRACTData collection, pre-processing, statistical analysis, feature extraction and classification are the main steps of Biomedical Image Processing. This paper presents a proposed model employing all the steps used for biomedical image processing. Different proteins were extracted from the heat map which is due to the combination of three input proteins was pre-processed. Statistical analysis is applied to all the different proteins to yield the marker proteins. ROI was extracted from each marker protein on which second order statistical transform method was applied to evaluate features. In this paper, authors have calculated the mean (observed and predicted), standard deviation, sum of squared error, absolute mean error using different classifiers like k-NN with different distance approaches and SVM using different kernels for selected marker proteins which provide better accuracy and least run-time complexity. Based on the best results, k-NN classifier using Euclidean distance and SVM classifiers using Linear Kernels yields an accuracy of 75.6% and 85.8%, respectively, that result better performance in comparison with other state of the art techniques.