Underground Characteristics Extraction Method Using the Fully Polarized Subsurface Penetrating Radar of Zhurong Rover

Robust speech recognition over telephone lines severely depends on the choice of the feature extraction and classification methods. In order to get the highest possible performance of the speech recognizer a number of commonly used feature extraction methods (MFCC, LPC, PLP, RASTA-PLP) and classification methods (MLP, LVQ, HMM) were tested on the same telephone speech data. All combinations of feature extraction and classification methods were computed and several parameters of both methods where changed in order to find a non-local maximum of recognition accuracy. The paper does not describe a comparison of classification but of feature extraction methods because it is clear that an HMM would outperform both LVQ and MLP. The big question is if the same feature extraction methods always lead to the best results, no matter which classifier is used!. >

Falling has become the leading cause of non-fatal and preventable injuries. The majority of existing fall detection systems (FDSs) rely on motion sensors. However, most of the applied feature extraction methods result in longer processing time and higher computation. This study proposes a novel attitude feature extraction (AFE) method that extract five feature sets from 54 raw measurements obtained from nine inertial measurement units (IMUs) placed on the firefighter’s protective clothing. Our results indicate that the metrics of attitude feature extraction (AFEM) method outperforms the existing metrics of raw feature extraction (RFEM) method in the fall detection. In addition, the proposed method reduces the algorithm processing time and computations significantly. This enables on-device fall detection classification on constrained processing architectures.

Extracting exact features from noisy point data is an important problem, in practice, for the application of reverse engineering. Several feature extraction methods have been used to handle noisy point data, such as the "angular" method and the "chordal" method. They work well for most cases, but the generation of extra features cannot be avoided for some cases. A new feature extraction method that deals with noisy scanned point data is proposed in this paper. We call it the iterative angular feature extraction (IAFE) method, since it extends the concept of the angular method. The IAFE method first distinguishes the feature regions from point clouds, then the iterative algorithm is applied to refine each feature region into ultimate feature points. A "noise dilution" concept is used to reduce the noise effect. A "multiple point" algorithm, an "angle variation" algorithm and an "iterations for convergence" algorithm are developed to implement the noise dilution concept. The IAFE(I) method for planar models and the IAFE(II) method for curved models are designed. The IAFE method demonstrated its usefulness in dealing with noisy point data.

The feature extraction is the most critical technology of text categorization. The method of feature extraction from Chinese text based on CILIN is different from the conventional feature extraction, which uses two feature extraction methods. This method is good at dealing with synonyms and polysemes, and reducing the dimension. Firstly, it uses the method of feature extraction from Chinese text based on CILIN to analyze the meaning of key words. Secondly, use the mutual information to extract the feature, it can give the relation between class and lemma. The experiment results proposed that comprehend to the meaning of key words can distinctively improve the text classification precision.

China's Mars probe including an orbiter and a landing rover will be launched by 2020. A subsurface penetrating radar (SPR) instrument has been selected to be a part of the payload on the rover. The main scientific objective of the SPR is to characterize the thickness and sub-layer distribution of the Martian soil. The SPR consists of two channels. The low frequency channel of the SPR will provide a penetration depth of 10 to 100 meters with a resolution of a few meters within the Martian soil. The higher frequency channel will penetrate to a depth of 3 to 10 meters with a resolution of a few centimeters within the Martian soil. The SPR first version prototype was designed and some field tests have been conducted with it.

ICD standardizes diagnosis codes globally, aiding payments, research, planning, and quality management. Its complexity leads to longer exams, higher training costs, increased workforce needs, coding errors, and unreliable data. Automated ICD systems using ML address these issues. Long medical notes complicate ML, making feature extraction crucial for efficient ICD classification. Despite numerous studies, no systematic analysis of feature extraction methods, especially in deep learning (DL), exists. The MIMIC-III dataset is used with two preprocessing combinations, fundamental and advanced. TF-IDF, word2vec, GloVe, fastText, and BERT feature extraction methods are compared using DL models such as NN, CNN, and BiLSTM. For word2vec and fastText, CBOW and skip-gram architectures are compared. ROC-AUC, F1-score, precision, and recall metrics are calculated for DL performances. Advanced preprocessing improves performance for all feature extraction and DL methods. The best results for advanced preprocessing are micro ROC-AUC of 91.74\% (BiLSTM+fastText (skip-gram)), macro ROC-AUC of 88.58\% (BiLSTM+word2vec (CBOW)), micro F1/precision of 64.84\%/62.34\% (BiLSTM+word2vec (CBOW)), micro recall of 68.16\% (BiLSTM+fastText (skip-gram)), macro F1/precision of 59.67\%/57.71\% (BiLSTM+word2vec (CBOW)), and macro recall of 63.38\% (BiLSTM+fastText (skip-gram)). FastText is the most successful feature extraction method in DL models with fundamental preprocessing. However, models using well-implemented preprocessing highlight other feature extraction methods that perform better and operate more quickly. As DL model performance improves, differences between feature extraction performances diminish. Though not focused on the best results, CNN and BiLSTM with word2vec, GloVe, and fastText are competitive with current studies. Lastly, if computing power is limited, CNN may be preferable over BiLSTM with these feature extraction methods.

The estimation of soil organic carbon based on visible near-infrared spectroscopy (VNIR) and hyperspectral image (HSI) has many advantages. However, the estimation accuracy has been a challenge that limits the wide application of spectral and hyperspectral imaging. Fully extracting the spectral and hyperspectral features of soil carbon information helps improve the estimation accuracy of soil organic carbon. Therefore, feature extraction is an important part of soil organic carbon estimation. This paper introduces the research on soil organic carbon prediction based on VNIR and HSI, the feature extraction methods, and the multi-feature fusion methods. The feature extraction methods introduce handcrafted feature extraction methods and deep learning feature extraction methods. Multi-feature fusion methods are divided into multi-feature fusion in handcrafted feature extraction methods and deep learning feature extraction methods. This paper also points out the future research direction and presents new ideas to improve the prediction of soil organic carbon. Soil organic carbon prediction based on VNIR and HSI, when combined with the multi-feature fusion method, is of great significance in extracting effective features and improving the prediction accuracy of soil organic carbon. It provides technical support for studying carbon cycling and carbon sinks, also guides the prediction of other soil properties.

In this paper, the methods of classifying the Uighur language feature extraction have been studied. According to the feature that the Uighur language belongs to adhesion, three experiments were designed to inspect the influence on the accuracy of text classification by using different methods of feature extraction, the first experiment was designed to inspect the accuracy of text classification in case of stem segmentation by using the traditional methods of feature extraction, such as DF,IG,MI,CHI. The results show that the best classification accuracy rate is 91.34% by the method of DF feature extraction, while the best accuracy rate is 88.03% in the second experiment by the method of CHI feature extraction in the case of stem that are not segmented. The third experiment uses combination of feature selection methods, such as DF+IG,DF+MI,DF+CHI, and the result show that the accuracy rate of classification is 93.57% by the method of DF+CHI feature selection, which shows that it is the best method in all experiments.

The motor imagery (MI) based brain-computer interface systems (BCIs) can help with new communication ways. A typical electroencephalography (EEG)-based BCI system consists of several components including signal acquisition, signal pre-processing, feature extraction and feature classification. This paper focuses on the feature extraction step and proposes to use a combination of different feature extraction and feature reduction methods. The research presented in the paper explores the methods of band power, time domain parameters, fast Fourier transform and channel variance for feature extraction. These methods are investigated by combining them in pairs. The application of two feature extraction methods increases the number of selected features that can be redundant or irrelevant. The utilization of too many features can lead to wrong classification results. Therefore, the methods of feature reduction have to be applied. The following feature reduction methods are investigated: principal component analysis, sequential forward selection, sequential backward selection, locality preserving projections and local Fisher discriminant analysis. The combination of the methods of fast Fourier transform, channel variance and principal component analysis performed the best among the combinations of methods. The obtained classification accuracy of the above-mentioned combination of the methods is much higher than that of the individual feature extraction method. The novelty of the approach is based on consolidated sequence of methods for feature extraction and feature reduction.

In order to reduce unnecessary stops and expensive downtime originating from clutch failure of construction equipment machines; adequate real time sensor data measured on the machine in combination with feature extraction and classification methods may be utilized.This paper presents a framework with feature extraction methods and an anomaly detection module combined with Case-Based Reasoning (CBR) for on-board clutch slippage detection and diagnosis in heavy duty equipment. The feature extraction methods used are Moving Average Square Value Filtering (MASVF) and a measure of the fourth order statistical properties of the signals implemented as continuous queries over data streams. The anomaly detection module has two components, the Gaussian Mixture Model (GMM) and the Logistics Regression classifier. CBR is a learning approach that classifies faults by creating a new solution for a new fault case from the solution of the previous fault cases. Through use of a data stream management system and continuous queries (CQs), the anomaly detection module continuously waits for a clutch slippage event detected by the feature extraction methods, the query returns a set of features, which activates the anomaly detection module. The first component of the anomaly detection module trains a GMM to extracted features while the second component uses a Logistic Regression classifier for classifying normal and anomalous data. When an anomaly is detected, the Case-Based diagnosis module is activated for fault severity estimation.

CADefender: Detection of unknown malicious AutoLISP computer-aided design files using designated feature extraction and machine learning methods

Recently, many electricity retailers have been aggregating consumer power demand information from smart meter infrastructure, and applications that utilize these data are widely studied. For example, the estimation of customers' background information using their power demand information has attracted much interest; this information can be utilized in the marketing field for background-targeted advertising. In order to utilize power demand information effectively, an appropriate feature extraction method must be applied. In this paper, appropriate data extraction methods specific to power demand information are proposed. In the experiment, power demand data for Kawasaki city were used, and 19 feature data were extracted using the proposed method. The utility of the extracted features was assessed through the performance of classification estimation for two background information types, family structure and floor space. The classification problems are solved by applying two typical machine-learning algorithms, the support vector machine and k-nearest neighbor. In particular, analysis of variance (ANOVA) was applied to the 19 feature data, which were ranked according to the F value. Then, the n (n = [1, 2, 19]) best feature data were used as the input step by step, and the score for each condition was computed to derive the best feature set. According to the results, some of the feature data were considered to be irrelevant, and the best feature data set was successfully selected. Furthermore, thee scores when raw data were input were also computed and compared with the scores when the best feature data set was used. As a result, the performance was better when using processed data instead of raw data.

Acquiring knowledge about users’ opinion and what they say regarding specific features within an app, constitutes a solid steppingstone for understanding their needs and concerns. App review utilization helps project management teams to identify threads and opportunities for app software maintenance, optimization and strategic marketing purposes. Nevertheless, app user review classification for identifying valuable gems of information for app software improvement, is a complex and multidimensional issue. It requires foresight and multiple combinations of sophisticated text pre-processing, feature extraction and machine learning methods to efficiently classify app reviews into specific topics. Against this backdrop, we propose a novel feature engineering classification schema that is capable to identify more efficiently and earlier terms-words within reviews that could be classified into specific topics. For this reason, we present a novel feature extraction method, the DEVMAX.DF combined with different machine learning algorithms to propose a solution in app review classification problems. One step further, a simulation of a real case scenario takes place to validate the effectiveness of the proposed classification schema into different apps. After multiple experiments, results indicate that the proposed schema outperforms other term extraction methods such as TF.IDF and χ2 to classify app reviews into topics. To this end, the paper contributes to the knowledge expansion of research and practitioners with the purpose to reinforce their decision-making process within the realm of app reviews utilization.

main advantage of biometric system in security mode is either in verification process or identification process for the persons. Iris recognition is one of the fast, accurate, reliable and secure biometric techniques for human identification. It provides automatic authentication of an individual based on the characteristics and unique features in iris structure. Thus the most important step in biometric system is the method of extract feature from pattern, especially in using Artificial Neural Networks (ANN) in the matching (recognition) process. There will be a close relationship between the type of network used and the method of extracting features. In this paper, three method of features extraction is tested using three types of GLCM based on number of angles for each type (2- Ang, 3-Ang, 4-Ang) as First Order Statistics (FOS) and 10 statistical measures as Second Order Statistics (SOS) for each type with three models of PNN, so as the model created is dependent on number of classes (20, 25, 30) in each model. Experimental results proved that third method (4ang-GLCM) of feature extraction with higher trained classes (30) had given best Recognition Rate with accuracy 94.43%. Thus, experimental results have been indicated to the efficiency of the proposed system in recognition accuracy in comparison with the previous methods.

Metro stray currents flowing into transformer-neutral points cause the high neutral DC and a transformer to operate in the DC bias state.Because neutral DC caused by stray current varies with time, the neutral DC value cannot be used as the only characteristic indicator to evaluate the DC bias risk level. Thus, unified characteristic extraction and assessment methods are proposed to evaluate the DC bias risk of a transformer caused by stray current, considering the signals of transformer-neutral DC and vibration. In the characteristic extraction method, the primary characteristics are obtained by comparing the magnitude and frequency distributions of transformer-neutral DC and vibration with and without metro stray current invasion. By analyzing the correlation coefficients, the final characteristics are obtained by clustering the primary characteristics with high correlation. Then, the magnitude and frequency characteristics are extracted and used as indicators to evaluate the DC bias risk. Moreover, to avoid the influence of manual experience on indicator weights, the entropy weight method (EWM) is used to establish the assessment model. Finally, the proposed methods are applied based on the neutral DC and vibration test data of a certain transformer. The results show that the characteristic indicators can be extracted, and the transformer DC bias risk can be evaluated by using the proposed methods.