Feature Vector Construction Research Articles

Research on feature vector construction and classification recognition of typical mine microseismic waveforms

In the realm of mining seismology, the identification of microseismic waveforms is a critical step for enhancing the precision of microseismic event localization. Addressing the complexities and diverse sources of noise in subterranean mining environments, this study has developed a novel methodology for the automated categorization and recognition of characteristic microseismic waveforms. Initially, a comprehensive database for mining-induced microseismic waveforms was established, catering to their non-stationary and nonlinear attributes. Based on the distinct features of different microseismic waveforms, this study introduces an approach for classification and identification that relies on waveform characteristics. A feature selection model, KNN-LOO (K-Nearest Neighbor - Leave-One-Out), was employed to optimize the feature vectors, in addition to the utilization of Principal Component Analysis (PCA) for dimensionality reduction of the feature vectors. Ultimately, an enhanced BT-SVM (Binary Tree-Support Vector Machine) model was developed to accomplish the automated classification and identification of mining-induced microseismic waveforms. Experimental results indicate that, compared to conventional SVM classification methods, the proposed methodology not only exhibits theoretical and mechanistic advancements but also significantly boosts performance in practical applications. Specifically, the average misclassification rate for each category of waveforms is reduced to less than 5%, and the overall accuracy has increased by 20%. This considerably elevates both the accuracy and speed of microseismic waveform classification and recognition, thereby providing valuable insights for monitoring and analyzing microseismic activities within mining operations as well as enhancing the safety protocols for mineral extraction.

A DCT probability histogram-based ROI features for content-based natural and medical image retrieval applications

The effectiveness of a Content-based Image Retrieval (CBIR) system depends on selecting representative and salient visual features. Capturing the salient image features for forming a feature vector is an essential part of the CBIR process, as it can significantly affect retrieval efficiency. Most images contain a visually prominent region that perceives maximum image semantic information, known as Regions-of-Interest (ROIs). This paper suggests a new CBIR scheme that employs a Graph-based Visual Saliency (GBVS) map to extract the ROI of the image. Further, block-level Discrete Cosine Transforms (DCT) and subsequent histograms are computed for the DC and significant AC coefficients of the extracted ROI image. In addition, a low-dimensional feature vector is formed from the modified histograms. Subsequently, six statistical features from the background scene have also been computed for the final feature vector construction. As a result, the final feature vector contains ROI and background information in their proper proportion to improve retrieval performance. Moreover, the performance of the proposed CBIR scheme has been evaluated using one medical, three natural, two objects, one texture, and one natural heterogeneous image database. All these databases collectively contain 53732 different images from 200 image classes. The outcomes of the experiments demonstrate substantial improvement in the retrieval process with some existing related schemes. At the same time, it also shows that the performance of the proposed scheme is robust concerning all kinds of image databases.

Classification of texts on emergency situations in Almaty

Text classification is a process that includes stages and approaches for the effective classification of texts that are diverse in their structure. In this article, machine learning algorithms are implemented, such as the support vector method, logistic regression, and the k nearest neighborhood method for classifying texts collected from emergency news sites in Almaty. During the experiment, a special role was played by the data collection stage, as well as their subsequent processing. Prior to the classification of the data set, preliminary data processing was performed, which includes such steps as the removal of stop words, tokenization, stemming, lemmatization, feature extraction, and the construction of feature vectors. The data was obtained by automated collection of information from open sources using a script. Experimental results show that the classifier based on logistic regression provides the best performance results compared to other types of algorithms. The performance indicators of each algorithm were obtained, which allows us to perform a comparative analysis between them.

Sequence Representations and Their Utility for Predicting Protein-Protein Interactions.

Protein-Protein Interactions (PPIs) are a crucial mechanism underpinning the function of the cell. So far, a wide range of machine-learning based methods have been proposed for predicting these relationships. Their success is heavily dependent on the construction of the underlying feature vectors, with most using a set of physico-chemical properties derived from the sequence. Few work directly with the sequence itself. In this paper, we explore the utility of sequence embeddings for predicting protein-protein interactions. We construct a protein pair feature vector by concatenating the embeddings of their constituent sequence. These feature vectors are then used as input to a binary classifier to make predictions. To learn sequence embeddings, we use two established Word2Vec based methods - Seq2Vec and BioVec - and we also introduce a novel feature construction method called SuperVecNW. The embeddings generated through SuperVecNW capture some network information in addition to the contextual information present in the sequences. We test the efficacy of our proposed approach on human and yeast PPI datasets and on three well-known networks: CD9, the Ras-Raf-Mek-Erk-Elk-Srf pathway, and a Wnt-related network. We demonstrate that low dimensional sequence embeddings provide better results than most alternative representations based on physico-chemical properties while offering a far simple approach to feature vector construction.

Calibration-Free Cross-Camera Target Association Using Interaction Spatiotemporal Consistency

In this paper, we propose a novel calibration-free cross-camera target association algorithm that aims to relate local visual data of the same object across cameras with overlapping FOVs. Unlike other methods using object's own characteristics, our approach makes full use of the interactions between objects and explores their spatiotemporal consistency in projection transformation to associate cameras. It has wider applicability in deployed overlapping multi-camera systems with unknown or rarely available calibration data, especially if there is a large perspective gap between cameras. Specifically, we first extract trajectory intersection which is one of the typical object-object interactive behaviors from each camera for feature vector construction. Then, based on the consistency of object-object interactions, we propose a multi-camera spatiotemporal alignment method via wide-domain cross-correlation analysis. It realizes time synchronization and spatial calibration of the multi-camera system simultaneously. After that, we introduce a cross-camera target association approach using aligned object-object interactions. The local data of the same target are successfully associated across cameras without any additional calibration. Extensive experimental evaluations on different databases verify the effectiveness and robustness of our proposed method.

Proximity Queries on Terrain Surface

Due to the advance of the geo-spatial positioning and the computer graphics technology, digital terrain data has become increasingly popular nowadays. Query processing on terrain data has attracted considerable attention from both the academic and the industry communities. Proximity queries such as the shortest path/distance query, k nearest/farthest neighbor query, and top- k closest/farthest pairs query are fundamental and important queries in the context of the terrain surfaces, and they have a lot of applications in Geographical Information System, 3D object feature vector construction, and 3D object data mining. In this article, we first study the most fundamental type of query, namely, shortest distance and path query, which is to find the shortest distance and path between two points of interest on the surface of the terrain. As observed by existing studies, computing the exact shortest distance/path is very expensive. Some existing studies proposed ϵ -approximate distance and path oracles, where ϵ is a non-negative real-valued error parameter. However, the best-known algorithm has a large oracle construction time, a large oracle size, and a large query time. Motivated by this, we propose a novel ϵ -approximate distance and path oracle called the S pace E fficient distance and path oracle (SE), which has a small oracle construction time, a small oracle size, and a small distance and path query time, thanks to its compactness of storing concise information about pairwise distances between any two points-of-interest. Then, we propose several algorithms for the k nearest/farthest neighbor and top- k closest/farthest pairs queries with the assistance of our distance and path oracle SE . Our experimental results show that the oracle construction time, the oracle size, and the distance and path query time of SE are up to two, three, and five orders of magnitude faster than the best-known algorithm, respectively. Besides, our algorithms for other proximity queries including k nearest/farthest neighbor queries and top- k closest/farthest pairs queries significantly outperform the state-of-the-art algorithms by up to two orders of magnitude.

LiDAR-Based Windshear Detection via Statistical Features

Windshear is a kind of microscale meteorological phenomenon which can cause danger to the landing and takeoff of aircrafts. Accurate windshear detection plays a crucial role in aviation safety. With the development of machine learning, several learning-based methods are proposed for windshear detection, i.e., windshear and non-windshear classification. To obtain accurate detection results, it is significant to extract features that can distinguish windshear and non-windshear properly from the obtained wind velocity data. In this paper, we mainly introduce two statistical indicators derived from the Doppler Light Detection and Ranging (LiDAR) observational wind velocity data by plan position illustrate (PPI) scans for windshear features construction. Besides the indicators directly derived from the wind velocity data, we also study the visual information from the corresponding conical images of wind velocity. Based on the proposed indicators, we construct three feature vectors for windshear and non-windshear classification. Inspired by the idea of multiple instance learning, the wind velocity data collected in the 4 minutes within the reported time spot are considered in the procedure of feature vector construction, which can reduce the possibility of windshear features missing. Both statistical methods and clustering methods are applied to evaluate the effectiveness of the proposed feature vectors. Numerical results show that the proposed feature vectors have good effect on windshear and non-windshear classification and can be used to provide more accurate windshear alerting to pilots in practice.

Direction Prediction of Traffic Flow in Vissim Simulation Based on K Nearest Neighbor Algorithm

In order to make short-term prediction of the direction of traffic flow in urban roads, a short-term prediction method of urban road travel time based on K nearest neighbor algorithm and vissim simulation is constructed. First, the intersection of Shiji Road and Yingbin Road was selected as the survey site, and the number of vehicles in each direction of each entrance lane of the intersection was investigated using manual counting, and the signal timing of each phase of the intersection was investigated. Input the survey data into the vissim simulation software to get the travel time of each entrance lane in each direction. Then build a vissim simulation traffic flow direction prediction model based on the KNN algorithm, including the construction of feature vectors, cross-validation methods to determine K values, and local estimation methods. The experimental results show that the average relative error between the predicted traffic flow direction and the actual traffic flow direction tends to 0.27. Due to the small amount of data, the prediction result is more accurate.

Recognition of multi-cognitive tasks from EEG signals using EMD methods

AbstractMental task classification (MTC), based on the electroencephalography (EEG) signals is a demanding brain–computer interface (BCI). It is independent of all types of muscular activity. MTC-based BCI systems are capable to identify cognitive activity of human. The success of BCI system depends upon the efficient feature representation from raw EEG signals for classification of mental activities. This paper mainly presents on a novel feature representation (formation of most informative features) of the EEG signal for the both, binary as well as multi MTC, using a combination of some statistical, uncertainty and memory- based coefficient. In this work, the feature formation is carried out in the two stages. In the first stage, the signal is split into different oscillatory functions with the help of three well-known empirical mode decomposition (EMD) algorithms, and a new set of eight parameters (features) are calculated from the oscillatory function in the second stage of feature vector construction. Support vector machine (SVM) is used to classify the feature vectors obtained corresponding to the different mental tasks. This study consists the problem formulation of two variants of MTC; two-class and multi-class MTC. The suggested scheme outperforms the existing work for the both types of mental tasks classification.

Topic2Labels: A framework to annotate and classify the social media data through LDA topics and deep learning models for crisis response

The abundant use of social media impacts every aspect of life, including crisis management. Disaster management needs real-time data to be used in machine learning and deep learning models to aid their decision making. Mostly the data that is newly generated from social media is unstructured and unlabeled. Current text classification models based on supervised deep learning models heavily rely on human-labeled data that very small size and imbalanced in the context of disasters, ultimately affecting the generalization of models. In this study, we propose Topic2labels (T2L) framework which provides an automated way of labeling the data through LDA (latent dirichlet allocation) topic modeling approach and utilize Bert (the bidirectional encoder representation from transformer) embeddings for construction of feature vector to be employed to classify the data contextually. Our framework consists of three layers. In the first layer, we adopt LDA to generate the topics from the data, and develop a new algorithm to rank the topics, and map the highest ranked dominant topic into label to annotate the data. In the second layer, we transform the labeled text into feature representation through Bert embeddings and in the third layer we leveraged deep learning models as classifiers to classify the textual data into multiple categories. Experimental results on crisis-related datasets show that our framework performs better in terms of classification performance and yields improvement as compared to other baseline approaches.

Subspace-based feature extraction on multi-physiological measurements of automobile drivers for distress recognition

The automotive industry has accelerated the utilization of Intelligent Transport Systems (ITS) in vehicles for increased driving safety. In this paper, a novel and well-done subspace feature extraction scheme on the physiological signals acquired by wearable sensors, for drivers’ distress level detection to be introduced as an ITS is proposed and verified on the publicly available MIT-BIH PhysioNet Multi-parameter Database. The proposed scheme includes two phases where time-domain statistical feature extraction is first realized on the electrocardiogram (ECG), hand galvanic skin response (hand GSR), foot galvanic skin response (foot GSR), electromyogram (EMG), and respiration (RESP) signals, and secondly subspace feature vector construction is appreciated by applying Discriminative Common Vector (DCV) decomposition on the statistical feature vectors. The distress levels of the drivers are determined as low, moderate, and high by utilizing both the statistical and the subspace feature vectors using Support Vector Machines (SVM) classifier by 2-fold cross-validation technique. A maximum of 88.89 % classification accuracy is achieved using statistical features in 7384 s while it is increased to 100 % in 3,421 s when subspace features are employed. The increased classification accuracy in decreased time consumption evidently shows the success of the proposed feature extraction scheme.

Identification of Lysine Carboxylation Sites in Proteins by Integrating Statistical Moments and Position Relative Features via General PseAAC

Background:Carboxylation is one of the most biologically important post-translational modifications and occurs on lysine, arginine, and glutamine residues of a protein. Among all these three, the covalent attachment of the carboxyl group with the lysine side chain is the most frequent and biologically important type of carboxylation. For studying such biological functions, it is essential to correctly determine the lysine sites sensitive to carboxylation.Objective:Herein, we present a computational model for the prediction of the carboxylysine site which is based on machine learning.Methods:Various position and composition relative features have been incorporated into the Pse- AAC for construction of feature vectors and a neural network is employed as a classifier. The model is validated by jackknife, cross-validation, self-consistency, and independent testing.Results:The results of the self-consistency test elaborated that model has 99.76% Acc, 99.76% Sp, 99.76% Sp, and 0.99 MCC. Using the jackknife method, prediction model validation gave 97.07% Acc, while for 10-fold cross-validation, prediction model validation gave 95.16% Acc.Conclusion:The results of independent dataset testing were 94.3% which illustrated that the proposed model has better performance as compared to the existing model PreLysCar; however, the accuracy can be improved further, in the future, due to the increasing number of carboxylysine sites in proteins.

Graph classification algorithm based on graph structure embedding

With the application of data mining in many fields such as information science, bioinformatics, and network intrusion detection, more and more data are showing new features such as strong structuration and complex relationships between data. As a complex data structure, a graph can be used to describe the relationship between things. Traditional graph classification methods based on graph feature vector construction need to select a feature vector construction criterion in advance, such as graph-based theoretical indicators or graph-based topology occurrences, and then extract features from each graph in the graph set according to the designated criterion. However, the construction method of the graph feature vector is easy to lose the graph structural information and requires strong professional knowledge. Inspired by the Word2Vec and Doc2Vec models in the Natural Language Processing (NLP), this paper first constructs a “word list” of graph data consisting of subgraphs. Then a neural network for training graph embedding is designed with the graph itself as its input, and the “word” in the graph and the attribute features of the graph are used as its output, so that the neural network automatically learns the graph embedding corresponding to each graph. The graph embedding not only reflects the features of the graph itself but also includes the relative relationship among graphs. Finally, on the basis of the well-trained graph embedding, the common classifier can be used to classify graphs. Based on real-world bioinformatics and social data sets, the experiments demonstrate that the proposed graph classification algorithm has advantages over the existing graph classification algorithms based on feature vector construction.

Low-Complexity Failed Element Diagnosis for Radar-Communication mmWave Antenna Array with Low SNR

The millimeter-wave (mmWave) antenna array plays an important role in the excellent performance of wireless sensors networks (WSN) or unmanned aerial vehicle (UAV) clusters. However, the array elements are easily damaged in its harsh working environment but hard to be repaired or exchanged timely, resulting in a serious decline in the beamforming performance. Thus, accurate self-diagnosis of the failed elements is of great importance. In previous studies, there are still significant difficulties for large-scale arrays under extremely low SNR. In this paper, a diagnosis algorithm with low complexity and high reliability for the failed elements is proposed, which is based on a joint decision of communication signal and sensing echoes. Compared with the previous studies, the complexity of the algorithm is reduced by the construction of low-dimensional feature vectors for classification, the decoupling of the degree of arrival (DOA) estimation and the failed pattern diagnosis, with the help of the sub-array division. Simulation results show that, under an ultra-low SNR of −12.5 dB for communication signals and −16 dB for sensing echoes, an accurate self-diagnosis with a block error rate lower than 8% can be realized. The study in this paper will effectively promote the long-term and reliable operation of the mmWave antenna array in WSN, UAV clusters and other similar fields.

An Improved Fault Diagnosis Method of Rotating Machinery Using Sensitive Features and RLS-BP Neural Network

An improved algorithm with feature selection and neural network classification is proposed in this paper to investigate the fault diagnosis problem of rotating machinery. The feature vectors are constructed by extracting the time- and frequency-domain characteristics of the overall machine under multiple operating conditions. To strengthen the fault diagnostic ability, an improved sensitive feature selection algorithm is proposed by improving the distance evaluation (DE) method and reconstructing a low-dimensional sensitive feature sample with selectively chosen parameters from multidimensional feature vectors. The recursive least square backpropagation (RLS-BP) neural network algorithm is used for fault diagnosis by classifying the feature vectors of normal signal and faulty signals. The effectiveness of the proposed method is verified via hardware experiments using wind turbine drivetrain diagnostics simulator (WTDDS) by comparing with conventional feature vector construction methods and neural network algorithm.

Information Theoretic Feature Transformation Learning for Brain Interfaces.

A variety of pattern analysis techniques for model training in brain interfaces exploit neural feature dimensionality reduction based on feature ranking and selection heuristics. In the light of broad evidence demonstrating the potential sub-optimality of ranking-based feature selection by any criterion, we propose to extend this focus with an information theoretic learning-driven feature transformation concept. We present a maximum mutual information linear transformation and a nonlinear transformation framework derived by a general definition of the feature transformation learning problem. Empirical assessments are performed based on electroencephalographic data recorded during a four class motor imagery brain-computer interface (BCI) task. Exploiting the state-of-the-art methods for initial feature vector construction, we compare the proposed approaches with conventional feature selection-based dimensionality reduction techniques, which are widely used in brain interfaces. Furthermore, for the multi-class problem, we present and exploit a hierarchical graphical model-based BCI decoding system. Both binary and multi-class decoding analyses demonstrate significantly better performances with the proposed methods. Information theoretic feature transformations are capable of tackling potential confounders of conventional approaches in various settings. We argue that this concept provides significant insights to extend the focus on feature selection heuristics to a broader definition of feature transformation learning in brain interfaces.

Robust and effective clothes recognition system based on fusion of Haralick and HOG features

In today's modern era, when the computer has become a necessity of an individual, shopping has shifted from shop to online shopping. This kind of clothes classification is used for knowing the name of the cloth that we have seen any movie, serial or anywhere else. In this paper, we present an efficient method to recognise the clothes in natural scenes as well as in the cluttered background. The proposed approach includes three phases: extraction of region of interest (ROI); construction of feature vector; classification. We have validated the proposed approach using our dataset which contains cluttered background images as well as on deep fashion standard dataset. The proposed method successfully resolved the issues of misclassification of clothes in the cluttered background with different illumination conditions. Experimental results show that the proposed technique successfully achieved 88.36% clothes recognition rate.

A Powerful Equalizer Based on Modified SVM Classifier Without Nonlinear Kernel Enabled 100-Gb/s NG-EPON System With 10-G Class

For the future high-speed next-generation Ethernet passive optical network (NG-EPON) system, the low-cost system could be kept by adopting low bandwidth devices. To compensate for the signal distortion and improve the performance of high-speed bandwidth limited system, we propose a special feature vector construction to achieve a new equalizer scheme based on support vector machines (SVM) classifier. For the first time, we mathematically analyze the model of the bandwidth-limited optical transmission system and theoretically derive the linear separability of the proposed feature vector formed linear hyperplane in 100-Gb/s NG-EPON system. Furthermore, 4 × 25-Gb/s non-return to zero (NRZ) and 2 × 50-Gb/s four-level pulse amplitude modulation (PAM-4) per wavelength transmission experiments with 10-G class optics system are demonstrated to prove the feasibility of our scheme. The experimental results show that, for 25-Gb/s/λ NRZ transmission, compared to the decision feedback equalizer (DFE) and feed-forward equalizer (FFE), the receiver sensitivity achieved by our scheme can be improved by about 0.5-dB @BER=3.8e-3 with limited training sequence. For the 50-Gb/s/λ PAM-4 case, the SVM-based equalization method improves the receiver sensitivity for about 1-dB.

Generalized halftone classification approach using stochastic analysis

Halftone classification is a primary requisite for the perfect reconstruction of binary patterns during inverse halftone process. Majority of the halftone classification techniques are either limited to error diffused halftone or to limited categories, and cannot be generalized to all halftone versions. In this scenario, a new classification approach is proposed based on the premise that the stochastic analysis can uniquely characterize the halftone patterns. The proposed scheme exploits inherent association of halftone patterns and stochastic geometry, and utilizes its spatial and spectral parameters for feature vector construction. Extreme learning machine based multi-classifier model is adopted resulting in rapid and accurate classification. A digital halftone database comprises of 96 reference images along with the 21 varieties of halftone and multitone class is developed. From extensive analysis, it has been validated that the proposed scheme can achieve a 100% classification rate on many halftone versions and exhibit superior performance over the exiting classification techniques.

Malicious Website Detection Based on URLs Static Features

Real-time and effectiveness are two basic requirements in URL-based malicious website detection. Based on the analysis of malicious URLs construction pattern, this paper puts forward a method to detect malicious website based on URLs static features. All features used in classification are extracted using statistic and there is no need to online access to obtain additional information about website, which reduces detection time greatly. At the same time, a feature reduction method is given in feature vector construction. Combined with reasonable selection of machine learning algorithm, our method can achieve both high accuracy and high efficiency. A series of comparative experiments proved the effectiveness of our method.