New Type Of Feature Research Articles

State of health estimation of lithium-ion batteries based on fine-tuning or rebuilding transfer learning strategies combined with new features mining

Accurate state of health (SOH) estimation of lithium-ion batteries is essential to ensure the reliability of power equipment. However, the degradation trajectory of different cells and different types of batteries is not repeatable. At present, there is no unified model or method to effectively predict SOH for all batteries. Therefore, a new SOH estimation method is proposed in the paper. Firstly, two types of new features are proposed in this paper. One is the voltage features extracted from the constant-current charging stage, and the other is the capacity recovery feature. They are used to reflect the nonlinear degradation process of the battery. Secondly, the relationship between features and SOH is established by using the LSTM model, which can prevent the problem of gradient vanishing and gradient explosion during model learning. Finally, for the inconsistencies between the same type or different types of batteries, two different transfer learning strategies (fine-tuning and rebuilding) are proposed in this paper, and the effectiveness of the proposed features and transfer learning strategies is verified on three open-source battery data sets (NASA, Oxford, and CALCE). Experimental results show that the SOH estimation method proposed in the paper has good universality, robustness, and accuracy.

Feature-based modeling for variable fractal geometry design integrated into CAD system

Fractal geometry has been widely applied in computer graphics to visually represent natural objects, which cannot be easily represented by Euclidean geometry. Other than visual representation, parametric fractal-like CAD models have been used in industry for the design and manufacture of products with self-similarity geometric features. However, contemporary CAD systems cannot construct CAD fractal models efficiently, as most CAD fractal models are still built manually by designers. The construction and modifications for such CAD fractal models are tedious and time-consuming.This paper aims to build and modify CAD fractal models more efficiently in CAD systems. A method for automatically modeling parametric fractal-like geometry objects in CAD systems is proposed. Feature-based modeling methods have been applied for realizing algorithm-based CAD fractal model generation. A new type of feature, called CAD fractal feature is proposed and implemented to bridge the gap between CAD and fractal geometry. The fractal-like CAD model is constructed iteratively with a dynamic generator for each iteration cycle. Design parameters are extracted as inputs for dynamic fractal generators, which allows users to define, parameterize, generate and edit a fractal design with variable parameters. The parameters of the fractal generator can also take different values for different model construction iterations, which makes the generated models not necessarily to be strictly fractal but can also be fractal-like. The iterative generation method (IGM) is proposed as a built-in method in a CAD fractal feature for modeling complex but self-similar geometric models in an objected-oriented approach with high efficiency and sufficient design flexibility. As a result, designers do not need advanced knowledge of fractal geometry to build and modify CAD models with self-similar features. An extruded Koch Snowflake CAD model with parametric geometric variants has been built as a case study for demonstration of the proposed modeling method.

Infant Cry Signal Diagnostic System Using Deep Learning and Fused Features.

Early diagnosis of medical conditions in infants is crucial for ensuring timely and effective treatment. However, infants are unable to verbalize their symptoms, making it difficult for healthcare professionals to accurately diagnose their conditions. Crying is often the only way for infants to communicate their needs and discomfort. In this paper, we propose a medical diagnostic system for interpreting infants' cry audio signals (CAS) using a combination of different audio domain features and deep learning (DL) algorithms. The proposed system utilizes a dataset of labeled audio signals from infants with specific pathologies. The dataset includes two infant pathologies with high mortality rates, neonatal respiratory distress syndrome (RDS), sepsis, and crying. The system employed the harmonic ratio (HR) as a prosodic feature, the Gammatone frequency cepstral coefficients (GFCCs) as a cepstral feature, and image-based features through the spectrogram which are extracted using a convolution neural network (CNN) pretrained model and fused with the other features to benefit multiple domains in improving the classification rate and the accuracy of the model. The different combination of the fused features is then fed into multiple machine learning algorithms including random forest (RF), support vector machine (SVM), and deep neural network (DNN) models. The evaluation of the system using the accuracy, precision, recall, F1-score, confusion matrix, and receiver operating characteristic (ROC) curve, showed promising results for the early diagnosis of medical conditions in infants based on the crying signals only, where the system achieved the highest accuracy of 97.50% using the combination of the spectrogram, HR, and GFCC through the deep learning process. The finding demonstrated the importance of fusing different audio features, especially the spectrogram, through the learning process rather than a simple concatenation and the use of deep learning algorithms in extracting sparsely represented features that can be used later on in the classification problem, which improves the separation between different infants' pathologies. The results outperformed the published benchmark paper by improving the classification problem to be multiclassification (RDS, sepsis, and healthy), investigating a new type of feature, which is the spectrogram, using a new feature fusion technique, which is fusion, through the learning process using the deep learning model.

Designing compact features for remote stroke rehabilitation monitoring using wearable accelerometers

Stroke is known as a major global health problem, and for stroke survivors it is key to monitor the recovery levels. However, traditional stroke rehabilitation assessment methods (such as the popular clinical assessment) can be subjective and expensive, and it is also less convenient for patients to visit clinics in a high frequency. To address this issue, in this work based on wearable sensing and machine learning techniques, we develop an automated system that can predict the assessment score in an objective manner. With wrist-worn sensors, accelerometer data is collected from 59 stroke survivors in free-living environments for a duration of 8 weeks, and we map the week-wise accelerometer data (3 days per week) to the assessment score by developing signal processing and predictive model pipeline. To achieve this, we propose two types of new features, which can encode the rehabilitation information from both paralysed and non-paralysed sides while suppressing the high-level noises such as irrelevant daily activities. Based on the proposed features, we further develop the longitudinal mixed-effects model with Gaussian process prior (LMGP), which can model the random effects caused by different subjects and time slots (during the 8 weeks). Comprehensive experiments are conducted to evaluate our system on both acute and chronic patients, and the promising results suggest its effectiveness.

Ventricular Fibrillation and Tachycardia Detection Using Features Derived from Topological Data Analysis

A rapid and accurate detection of ventricular arrhythmias is essential to take appropriate therapeutic actions when cardiac arrhythmias occur. Furthermore, the accurate discrimination between arrhythmias is also important, provided that the required shocking therapy would not be the same. In this work, the main novelty is the use of the mathematical method known as Topological Data Analysis (TDA) to generate new types of features which can contribute to the improvement of the detection and classification performance of cardiac arrhythmias such as Ventricular Fibrillation (VF) and Ventricular Tachycardia (VT). The electrocardiographic (ECG) signals used for this evaluation were obtained from the standard MIT-BIH and AHA databases. Two input data to the classify are evaluated: TDA features, and Persistence Diagram Image (PDI). Using the reduced TDA-obtained features, a high average accuracy near 99% was observed when discriminating four types of rhythms (98.68% to VF; 99.05% to VT; 98.76% to normal sinus; and 99.09% to Other rhythms) with specificity values higher than 97.16% in all cases. In addition, a higher accuracy of 99.51% was obtained when discriminating between shockable (VT/VF) and non-shockable rhythms (99.03% sensitivity and 99.67% specificity). These results show that the use of TDA-derived geometric features, combined in this case this the k-Nearest Neighbor (kNN) classifier, raises the classification performance above results in previous works. Considering that these results have been achieved without preselection of ECG episodes, it can be concluded that these features may be successfully introduced in Automated External Defibrillation (AED) and Implantable Cardioverter Defibrillation (ICD) therapies.

Using ECG signals for hypotensive episode prediction in trauma patients

Background and objectives:Bleeding is the leading cause of death among trauma patients both in military and civilian scenarios, and it is also the most common cause of preventable death. Identifying a casualty who suffers from an internal bleeding and may deteriorate rapidly and develop hemorrhagic shock and multiorgan failure is a profound challenge. Blood pressure and heart rate are the main vital signs used nowadays for the casualty clinical evaluation in the battlefield and in intensive care unit. However, these vitals tend to deteriorate at a relatively late stage, when the ability to prevent hazardous complications is limited. Identifying, treating, and rapidly evacuating such casualties might mitigate these complications. In this work, we try to improve a state-of-the-art method for early identification of Hypotensive Episode (HE), by adding electrocardiogram signals to several vital signs.Methods:In this research, we propose to extend the state-of-the-art HE early detection method, In-Window Segmentation (InWise), by adding new types of features extracted from ECG signals. The new predictive features can be extracted from ECG signals both manually and automatically by a convolutional auto-encoder. In addition to InWise, we are trying to predict HE using a Transformer model. The Transformer is using the encoder output as an embedding of the ECG signal. The proposed approach is evaluated on trauma patients data from the MIMIC III database.Results:We evaluated the InWise prediction algorithm using four different groups of features. The first feature group contains the 93 original features extracted from vital signs. The second group contains, in addition to the original features, 24 features extracted manually from ECG signal (117 features in total). The third group contains the original features and 20 ECG features extracted by the AE (113 features in total), and the last group is the union of all three previous groups containing 137 features. The results show that each model, which has used ECG data, is outperforming the original InWise model, in terms of AUC and sensitivity with p-value <0.001 (by 0.7% in AUC and up to 3.8% in sensitivity). The model which has used all three feature types (vital signs, manual ECG and AE ECG), outperforms the original model both in terms of accuracy and specificity with p-value <0.001 (by 0.3% and 0.4% respectively).Conclusion:The results show an improvement in the prediction success rates as a result of using ECG-based features. The importance of ECG features was confirmed by the feature importance analysis.

An ensemble approach to predict binding hotspots in protein-RNA interactions based on SMOTE data balancing and Random Grouping feature selection strategies.

The identification of binding hotspots in protein-RNA interactions is crucial for understanding their potential recognition mechanisms and drug design. The experimental methods have many limitations, since they are usually time-consuming and labor-intensive. Thus, developing an effective and efficient theoretical method is urgently needed. Here, we present SREPRHot, a method to predict hotspots, defined as the residues whose mutation to alanine generate a binding free energy change ≥2.0 kcal/mol, while others use a cutoff of 1.0 kcal/mol to obtain balanced datasets. To deal with the dataset imbalance, Synthetic Minority Over-sampling Technique (SMOTE) is utilized to generate minority samples to achieve a dataset balance. Additionally, besides conventional features, we use two types of new features, residue interface propensity previously developed by us, and topological features obtained using node-weighted networks, and propose an effective Random Grouping feature selection strategy combined with a two-step method to determine an optimal feature set. Finally, a stacking ensemble classifier is adopted to build our model. The results show SREPRHot achieves a good performance with SEN, MCC and AUC of 0.900, 0.557 and 0.829 on the independent testing dataset. The comparison study indicates SREPRHot shows a promising performance. The source code is available at https://github.com/ChunhuaLiLab/SREPRHot. Supplementary data are available at Bioinformatics online.

Segments Interpolation Extractor for Finding the Best Fit Line in Arabic Offline Handwriting Recognition Words

In the last few years, deep learning-based models have made significant inroads into the field of handwriting recognition. However, deep learning requires the availability of massive labelled data and considerable computation for training or automatic feature extraction. The role of handcrafted features and their significance is still crucial for a specific language type because it is a unique way of writing the characters. These are primitive segments that describe the letter horizontally or vertically distinguish an Arabic letter. This article develops a new type of feature for handwriting using Segments Interpolation (SI) to find the best fitting line in each of the windows and build a model for finding the best operating point window size for SI features. The experimental design was done on two subsets of the Institute for Communications Technology/Ecole Nationale d’Ingenieurs de Tunis (IFN/ENIT) database. The first one contains 10 classes (C10), and the second one has 22 classes (C22). The extracted features were trained with Support Vector Machine (SVM) and Extreme Learning Machine (ELM) with different kernels and activation functions. The evaluation metrics from a classification perspective (Accuracy, Precision, Recall and F-measure) were applied. As a result, SI shows significant results with SVM 90.10% accuracy for C10 and 88.53% accuracy for C22.

Ensemble deep learning for tuberculosis detection using chest X-ray and canny edge detected images

Tuberculosis (TB) is a disease caused by Mycobacterium Tuberculosis. Detection of TB at an early stage reduces mortality. Early stage TB is usually diagnosed using chest x-ray inspection. Since TB and lung cancer mimic each other, it is a challenge for the radiologist to avoid misdiagnosis. This paper presents an ensemble deep learning for TB detection using chest x-ray and Canny edge detected images. This method introduces a new type of feature for the TB detection classifiers, thereby increasing the diversity of errors of the base classifiers. The first set of features were extracted from the original x-ray images, while the second set of features were extracted from the edge detected image. To evaluate the proposed approach, two publicly available datasets were used. The results show that the proposed ensemble method produced the best accuracy of 89.77%, sensitivity of 90.91% and specificity of 88.64%. This indicates that using different types of features extracted from different types of images can improve the detection rate.

A Reconstructed Indigenous Religious Tradition in Latvia

In the early 20th century, Dievturība, a reconstructed form of paganism, laid claim to the status of an indigenous religious tradition in Latvia. Having experienced various changes over the course of the century, Dievturība has not disappeared from the Latvian cultural space and gained new manifestations with an increase in attempts to strengthen indigenous identity as a result of the pressures of globalization. This article provides a historical analytical overview about the conditions that have determined the reconstruction of the indigenous Latvian religious tradition in the early 20th century, how its form changed in the late 20th century and the types of new features it has acquired nowadays. The beginnings of the Dievturi movement show how dynamic the relationship has been between indigeneity and nationalism: indigenous, cultural and ethnic roots were put forward as the criteria of authenticity for reconstructed paganism, and they fitted in perfectly with nativist discourse, which is based on the conviction that a nation’s ethnic composition must correspond with the state’s titular nation. With the weakening of the Soviet regime, attempts emerged amongst folklore groups to revive ancient Latvian traditions, including religious rituals as well. Distancing itself from the folk tradition preservation movement, Dievturība nowadays nonetheless strives to identify itself as a Latvian lifestyle movement and emphasizes that it represents an ethnic religion which is the people’s spiritual foundation and a part of intangible cultural heritage. In the 21st century, Dievturība is characterized by conflicting aspects: on the one hand, a desire is expressed to contrast itself and its ethnic views from globalization tendencies in its activities, but on the other hand New Age concepts and a self-reflexive character has entered its discourse.

Alzheimer's Disease Diagnosis and Severity Level Detection Based on Electroencephalography Modulation Spectral "Patch" Features.

Over the last two decades, electroencephalography (EEG) has emerged as a reliable tool for the diagnosis of cortical disorders such as Alzheimer's disease (AD). Typically, resting-state EEG (rsEEG) signals have been used, and traditional frequency bands (delta, theta, alpha, beta and gamma) have been explored. Recent studies, however, have suggested that non-conventional bands may lead to improved diagnostic performance. In this work, we propose a new type of features derived from the 2-dimensional modulation spectral domain representation of the rsEEG signal in order to characterize the neuromodulatory deficit emergent with AD. The proposed features are computed as the power in specific "patches" or regions of interest in the power modulation spectrogram, which are shown to be highly discriminant of AD severity levels. The proposed features were compared with traditional features used in the rsEEG AD monitoring literature. Results showed the proposed features not only achieving improved performance at discriminating between healthy normal elderly controls (Nold) and AD patients with varying severity levels, but also at monitoring severity levels (i.e., mild AD versus moderate AD). Moreover, the proposed features were shown to outperform traditional rsEEG features. Finally, we validated the biological origin of the proposed features by using source localization and comparing the obtained results with ones reported in the AD literature.

Imbalance learning for the prediction of N6-Methylation sites in mRNAs

BackgroundN6-methyladenosine (m6A) is an important epigenetic modification which plays various roles in mRNA metabolism and embryogenesis directly related to human diseases. To identify m6A in a large scale, machine learning methods have been developed to make predictions on m6A sites. However, there are two main drawbacks of these methods. The first is the inadequate learning of the imbalanced m6A samples which are much less than the non-m6A samples, by their balanced learning approaches. Second, the features used by these methods are not outstanding to represent m6A sequence characteristics.ResultsWe propose to use cost-sensitive learning ideas to resolve the imbalance data issues in the human mRNA m6A prediction problem. This cost-sensitive approach applies to the entire imbalanced dataset, without random equal-size selection of negative samples, for an adequate learning. Along with site location and entropy features, top-ranked positions with the highest single nucleotide polymorphism specificity in the window sequences are taken as new features in our imbalance learning. On an independent dataset, our overall prediction performance is much superior to the existing predictors. Our method shows stronger robustness against the imbalance changes in the tests on 9 datasets whose imbalance ratios range from 1:1 to 9:1. Our method also outperforms the existing predictors on 1226 individual transcripts. It is found that the new types of features are indeed of high significance in the m6A prediction. The case studies on gene c-Jun and CBFB demonstrate the detailed prediction capacity to improve the prediction performance.ConclusionThe proposed cost-sensitive model and the new features are useful in human mRNA m6A prediction. Our method achieves better correctness and robustness than the existing predictors in independent test and case studies. The results suggest that imbalance learning is promising to improve the performance of m6A prediction.

Adaptive virtual metrology for semiconductor chemical mechanical planarization process using GMDH-type polynomial neural networks

Virtual metrology (VM) is drawing more and more attention in the recent research of wafer to wafer control for semiconductor manufacturing. Although many different approaches for VM have been proposed, the adaptiveness of these approaches in feature selection and model complexity selection is still less discussed and most of the current researches rely on the summary statistics of the trace signals. In this work, an adaptive methodology based on the group method of data handling (GMDH) type polynomial neural networks is proposed to address these issues. In the proposed methodology, the processes for model selection and feature selection are fully automatic, and enhanced model performance can be achieved by employing two new types of features. To show the effectiveness of the propose methodology, the dataset from prognostics and health management (PHM) data challenge 2016 is employed to predict the material removal rate for the chemical-mechanical planarization process in semiconductor fabrication. The validation results report improved accuracy in comparison with several candidate methods, and the successful application of the proposed method suggests that the proposed method can be an effective tool for the virtual metrology in semiconductor manufacturing.

Semantics-aware Recommender Systems exploiting Linked Open Data and graph-based features

The recent spread of Linked Open Data (LOD) fueled the research in the area of Recommender Systems, since the (semantic) data points available in the LOD cloud can be exploited to improve the performance of recommendation algorithms by enriching item representations with new and relevant features.In this article we investigate the impact of the features gathered from the LOD cloud on a hybrid recommendation framework based on three classification algorithms, Random Forests, Naïve Bayes and Logistic Regression. Specifically, we extend the representation of the items by introducing two new types of features: LOD-based features, structured data extracted from the LOD cloud, as the genre of a movie or the writer of a book, and graph-based features, computed on the ground of the topological characteristics of both the bipartite graph-based representation connecting users and items, and the tripartite representation connecting users, items and properties in the LOD cloud.In the experimental session we assess the effectiveness of these novel features; results show that the use of information coming from the LOD cloud could improve the overall accuracy of our recommendation framework. Finally, our approach outperform several state-of-the-art recommendation techniques, thus confirming the insights behind this research.

Estimating effectiveness of twitter messages with a personalized machine learning approach

To improve a tweet in Twitter, we would like to estimate the effectiveness of a draft before it is sent. The total number of retweets of a tweet can be considered as a measure for the tweet’s effectiveness. To estimate the number of retweets for an author, we propose a procedure to learn a personalized model from his/her past tweets. We propose three types of new features based on the contents of the tweets: Entity, Pair, and Topic. Empirical results from seven authors indicate that the Pair and Topic features have statistically significant improvements on the correlation coefficient between the estimates and the actual numbers of retweets. We study different combinations of the three types of features, and many of the combinations significantly improve the result further.

Sparse coding of 2D-slice Zernike moments for SAR ATR

ABSTRACTIn this article, a new type of feature, named two-dimensional (2D)-slice Zernike moments, is proposed for synthetic aperture radar (SAR) automatic target recognition (ATR). Target features play an extremely important role in the ATR system. Pixels with different scattering intensities distribute in different positions in SAR images, which represent target inherent signatures determined by the target’s characteristics, including global structure and local details. To extract these various scattering signatures, we developed a feature extraction technique named 2D-slice Zernike moments (2DS-ZMS), which can capture target global and local scattering field distribution information. First, the multilayer 2D-slices of a SAR image are extracted by uniformly cutting the 3D display SAR image along the amplitude direction. Then Zernike moments of each 2D-slice are calculated. Finally, the 2DS-ZMS of the SAR image are formulated into a column vector, called the feature vector. The obtained feature vectors of the targets are fed into a newly developed classifier, i.e. the sparse representation-based classifier (SRC). By projecting the testing sample feature vector on the dictionary made up of training samples feature vectors, the sparse coefficients are solved. The minimum reconstruction residual is adopted as the judgement criterion for predicting the test sample’s class label. Experiments on the moving and stationary target acquisition and recognition (MSTAR) data set validate the effectiveness and superiority of the proposed method.

Symmetrical feature for interpreting motor imagery EEG signals in the brain–computer interface

Feature extraction is an important issue of brain–computer interface (BCI). It determines whether the classification performance is high or low. In this paper, a new type of feature called “symmetrical feature” is proposed. This innovative feature extraction method is built upon the features “common spatial pattern (CSP)” algorithm. After an electroencephalographic signal is enhanced, class discrimination using the CSP algorithm can be extracted using optimal symmetrical axis chosen by a 10-fold cross-validation technique. Simulation results from nine data sets provided by brain–computer interface competition III and Iva showed that, on average, the proposed symmetrical feature can be combined with the CSP power band feature to boost the performance of the classification in a BCI system.

Improving Question Classification by Feature Extraction and Selection

Question classification is the task of predicting the entity type of the answering sentence for a given question in natural language. It plays an important role in finding or constructing accurate answers and therefore helps to improve quality of automated question answering systems. Different lexical, syntactical and semantic features was extracted automatically from a question to serve the classification in previous studies. However, combining all those features doesn't always give the best results for all types of questions. Different from previous studies, this paper focuses on the problem of how to extract and select efficient features adapting to each different types of question. We first propose a method of using a feature selection algorithm to determine appropriate features corresponding to different question types. Secondly, we design a new type of features, which is based on question patterns. We tested our proposed approach on the benchmark dataset TREC and using Support Vector Machines (SVM) for the classification algorithm. The experiment shows obtained results with the accuracies of 95.2% and 91.6% for coarse grain and fine grain data sets respectively, which are much better in comparison with the previous studies.

Augmenting bag-of-words: a robust contextual representation of spatiotemporal interest points for action recognition

Although traditional bag-of-words model, together with local spatiotemporal features, has shown promising results for human action recognition, it ignores all structural information of features, which carries important information of motion structures in videos. Recent methods usually characterize the relationship of quantized spatiotemporal features to overcome this drawback. However, the propagation of quantization error leads to an unreliable representation. To alleviate the propagation of quantization error, we present a coding method, which considers not only the spatial similarity but also the reconstruction ability of visual words after giving a probabilistic interpretation of coding coefficients. Based on our coding method, a new type of feature called cumulative probability histogram is proposed to robustly characterize contextual structural information around interest points, which are extracted from multi-layered contexts and assumed to be complementary to local spatiotemporal features. The proposed method is verified on four benchmark datasets. Experiment results show that our method can achieve better performance than previous methods in action recognition.

Diagnosis of Heart Valve Disorders through Trapezoidal Features and Hybrid Classifier

Numerous studies are being conducted in recent years focusing on phonocardiographic (PCG) signals due to their capability to characterize heart sounds. These characteristics can be exploited in developing computer-aided auscultation system as a complementary tool for clinicians in diagnosis of cardiovascular disorders. This study proposes a new type of features to distinguish five categories of heart sounds, including normal, mitral stenosis, mitral regurgitation,aortic stenosis, and aortic regurgitation. PCG signals were collected from online resources and training CDs. Wavelet packet transform was utilized for heart sound analysis as opposed to discrete wavelet transform that has been extensively used in the previous studies. Then, trapezoidal function was calculated for deriving feature vectors. A hybrid classifier was designed composing of three types of classifiers, multilayer perceptron (MLP) artificial neural network, k-nearest neighbor (KNN), and support vector machine (SVM), to classify feature vectors.The promising results demonstrate the effectiveness of the proposed trapezoidal features and hybrid classifier for heart sound classification.