Search

Automatic Modulation Classification (AMC) is an essential technology that is widely applied into various communications scenarios. In recent years, many Machine Learning and Deep-Learning methods have been introduced into AMC, and a lot of them apply different approaches to eliminate interference in complex Multiple-Input and Multiple-Output (MIMO) signals and improve classification performance. However, in practical communication systems, the perfect elimination of MIMO signal interference is impossible, and therefore classification performance suffers. In this paper, we propose a new AMC algorithm for MIMO system based on mutual information (MI) features extraction, which does not require a large amount of training data and the elimination of MIMO signal interference. In this approach, features based on mutual information are extracted using In-Phase and Quadrature (IQ) constellation diagrams of MIMO signals, which have not been explored previously. Our method can be effective since mutual information considers the interdependencies among variables and measures how much information about one variable reduces uncertainty about another, providing a valuable perspective for extracting higher-level and interesting features from the data. The effectiveness of our method is evaluated on several model and real-world datasets, and its applicability is proven.

The spectral images provide the dynamic characteristics of the voice signal in the time and frequency domains. However, extracting the predominant spectral features from the voice samples is still challenging. This work generates cochleagram images to unveil detailed spectral content of the voice samples to recognize dysphonic voice. Both sustained vowel (‘/a/’) and sentence voice samples are considered to include phonation, respiration, and resonance of the vocal tone. Also, gender bias is eliminated by considering male and female voice samples separately, as they have structurally different vocal tracts, pharynx, and oral cavities. The simulation results show that the cochleagram, coined with a designed pre-trained convolutional neural network (CNN), can achieve 95% accuracy in identifying dysphonic voices with sentence samples. A robust, noninvasive, and automated voice pathology detection system is effectively generated through perceptual analysis of voice signals. The proposed automated pathological voice detection system can objectively correlate the clinical findings and assist in monitoring the treatment progress of dysphonic voice on top of subjective assessment by clinicians.

Malware targeting user privacy has seen a surge in recent times, attributed to evolving global regulations and the boost of electronic commerce and online services. Moreover, recognizing privacy malware that employs obfuscation as evasion mechanism presents a major challenge due to its dynamics, resilience, and polymorphism at runtime, necessitating the application of forensic techniques such as memory dumping analysis in order to reveal suitable patterns and behaviors that enable its subsequent detection and classification. In this paper, we present three obfuscated privacy malware classifiers trained on the CIC-MalMem-2022 dataset. These solutions include a binary classifier to distinguish benign from malicious samples using logistic regression (LR), a multiclass classifier that further categorizes benign, spyware, ransomware, and trojan horse obfuscated privacy malware; and a more detailed multiclass classifier capable of discriminating benign samples from fifteen specific obfuscated privacy malware families. Multiclass classifiers were built using several traditional machine learning algorithms and a novel Deep Neural Network (DNN). We applied the Synthetic Minority Oversampling Technique (SMOTE) to address data imbalances. In particular, our results demonstrate that DNN outperforms traditional machine learning algorithms, yielding statistically significant improvements in key metrics. These achievements reach practical thresholds, suggesting the potential for enhanced malware protection systems. The dataset and all the coding files required for experiments reproducibility are publicly available at https://github.com/dcevallossalas/PrivacyMalwareClassifiers.

Customer reviews about a brand or product, movie reviews, and social media reviews can be analyzed through sentiment analysis. Sentiment analysis is used to identify the emotional tone of language to comprehend the attitudes, opinions, and feelings represented in online reviews. As for large data, it is a task that can take a lot of time and can be automated as the machine learns through the training and testing of data. Previously, various standard machine learning and deep learning models namely Recurrent Neural Network (RNN), Convolutional Neural Network (CNN), Long Short Term Memory (LSTM), Naïve Bayes (NB), Support Vector Machine (SVM), Gated Recurrent Unit (GRU) have been used. The key issue in our research is that when text is provided to LSTM directly, it cannot adequately extract informative features from the text, leading to less accurate findings. The softmax layer of Stacked Auto-encoder when used directly to categorize the extracted features, is power-constrained and unable to do so accurately. A hybrid of the Stacked Auto-encoder (SAE) and LSTM models was proposed. SAE is used for the extraction of relevant informative features. LSTM was used for further classification of sentiments based on the extracted features. The proposed model is evaluated on an IMDB dataset by splitting it into five different training testing ratios using the following performance evaluation metrics: confusion matrix, classification accuracy, precision, recall, sensitivity, specificity, and F1 score. The hybrid results performed best at a ratio of 90/10 and classified sentiments with an accuracy of 87%. The accuracy of proposed hybrid model is better than that of standard models namely RNN, CNN, LSTM, NB, SVM, and GRU.

Analyzing the decision of different genders during shopping is an interesting topic in the neuromarketing industry. This study explores the EEG signal acquisition of twenty subjects in response to Color and Black-White (CL/BW) images and analysis both linear and nonlinear features in different brain regions. The Wilcoxon Rank Sum statistical test was utilized to determine the significance of features in identifying discriminative channels and frequency bands. The results show that the activation of different bands and regions are dependent on the subject’s preference and the color of stimuli which are evaluated by spectral scalp map and power spectral density analysis on the the regions. Then, random forest, support vector machine, k-nearest neighbors, and linear discriminant analysis classifiers were also employed to identify the most significant active regions of different brain lobes in both human genders. For the female group in the Like task with CL/BW stimuli images, the classification accuracy significantly increased to 96.47% by combining all frequency bands in the random forest algorithm. On the other hand, for the male group, using the gamma frequency band in the k-nearest neighbors algorithm led to a notable improvement in classification accuracy, reaching 95.32% for the Like task with CL/BW stimuli images. The study provides insights into the influence of black-white colors on marketing products and neuromarketing strategies. The research also revealed differences in the time taken for males and females to make Like and Dislike decisions, as well as the decision-making time for Like and Dislike CL/BW images. The female group took approximately 2.5 seconds to choose an image of a product, whilst the male group took 2.5-3 seconds. The study’s findings have significant implications for the field of neuromarketing, emphasizing the importance of careful stimulus selection and classifier choice for classification tasks.

Driver distraction is one of the primary causes of crashes. As a result, there is a great need to continuously observe driver state and provide appropriate interventions to distracted drivers. Cognitive distraction refers to the “look but not see” situations when the drivers’ eyes are focused on the forward roadway, but their mind is not. Typically, cognitive distractions can result from fatigue, conversation with a co-passenger, listening to the radio, or other similarly loading secondary tasks that do not necessarily take a driver’s eyes off the roadway. This makes it one of the hardest distractions to detect as there are no visible clues of driver distraction. In this study, we have identified features from different sources including eye-tracking, physiological, and vehicle kinematics data that are relevant towards the classification of distracted and non-distracted drivers via the analysis of data collected from a driving simulator study involving 40 drivers across multiple driving scenarios. The key classification algorithms implemented include Random Forest, Decision Trees and Support Vector Machines. A reduced feature set including pupil area, pupil vertical and horizontal motion was found to be predictive of driver distraction while maintaining an average accuracy of 90% across various road types. Additionally, the impact of road types on driver behaviour was also identified. The findings of the study has practical application towards the design of driver distraction monitoring systems.

The grid reinforcement, advanced grid stabilizing systems, and inverter-interfaced loads have varied power system dynamics. The changing trends of various dynamic phenomena need to be scrutinized to ensure future grid reliability. A dynamic behavior-based event signature library of phasor measurement unit (PMU) data has great potential to discover new and unprecedented event signatures. This paper presents an event signature library design that further defines more granular event categories within the major event categories (e.g., frequency, voltage, and oscillation events) provided by electric utilities and regional transmission organizations. The proposed library design embraces a supervised machine learning approach with a deep neural network (DNN) model and manually-generated labels. The input of the model uses representative PMUs that evidently express dominant event signatures. The performance of the event categorization module was evaluated, via information entropy, against labels generated automatically from clustering analyses. We applied the event signature library design to two years of over 1000 actual events in the bulk U.S. power system. The module obtains remarkable event discrimination capability.

Sepsis is a lethal infection-related illness that has an extremely high fatality rate, especially among intensive care unit patients. Early and precise recognition of sepsis is critical as delayed treatment increases the mortality rate dramatically. System inflammatory response syndrome, quick sequential organ failure assessment, and modified early warning score are the traditional clinical score systems in practice to detect sepsis. But the scoring systems fail in the early prediction of sepsis, a stage in which if a patient is treated immediately, the mortality rate will reduce significantly. The proposed classifier can accurately predict sepsis up to six hours before the disease is clinically diagnosed. The patient’s electronic medical records, demographics, and vital signs are used to achieve this. The study uses data set adaptive data preprocessing strategies. The proposed method adds value to existing literature by introducing a novel outlier-based mean-median data imputation technique that enhances the prediction’s overall accuracy. The primary factors that influence the classifier’s predictions have been outlined, making the model easier to understand for medical professionals. For the classification of patients as sepsis positive or negative, four algorithms were investigated: Random Forest, Logistic Regression, Gradient Boosting, and Decision Tree. Of all the prediction algorithms, Random Forest gives the best results with an accuracy of 99.01%, F1-score of 99%, and an area under the receiver operator characteristic curve of 99.99%. Even for a 24-hour early prediction of sepsis, the random forest method is proven to provide greater prediction accuracy while logistic regression provides the least prediction accuracy. We attribute this to the fact that, unlike regression models, random forests do not require that the model have a linear relationship between the dependent and independent variables. The evaluation measures produced are useful and can be tremendously valuable in predicting sepsis in a timely and accurate manner.

Among pets, dogs are very famous in the whole world. The owners of dogs are very cautious about the well-being of their dogs. The well-being of dogs can be ensured by continuous monitoring of their activities. Studies related to activity detection have gained much popularity due to the significant progress in sensor technology during the last few years. Automatic monitoring of pet applications includes real-time monitoring systems and surveillance which detect the pets with high accuracy using the latest pet activity classification techniques. The revolution in the domain of technology has allowed us to obtain better results using latest techniques. Convolutional neural networks (CNNs) 1D recently become a cutting-edge approach for signal processing-based systems such as patient-individual ECG categorization, sensor-based health monitoring systems, and anomaly identification in manufacturing areas. Adaptive and compact 1D models have several advantages over their conventional 2D counterparts. A limited dataset is sufficient to train a 1D CNN efficiently while 2D CNNs require a plethora of data for training. Its architecture is not very complicated, so it is suitable for real-time detection of activities. The main goal of this study is to develop a state-of-the-art system that can detect and classify the activities based on sensors’ data (accelerometer, and gyroscope. We proposed a 1D CNN-based system for pet activity detection. The objective of this study was to recognize ten pet activities such as walking, sitting, down, staying, eating, sideway, jumping, running, shaking, and nose work respectively, using wearable sensor devices based on deep learning technique. The data collection procedure for this study was conducted with 10 dogs of different breeds, sex (male=7, female = 3), age (age = 4±3), and sizes (small, medium, large) in a healthy environment. After collecting the data, the following steps, namely data synchronization, and data preprocessing were considered to remove the irrelevant data from the dataset. To overcome imbalanced problems in the dataset we used the class-weight technique. Subsequently, we applied 1D CNN algorithm using the class-weight technique. The model with the class-weight technique showed 99.70% training accuracy and 96.85% validation accuracy. The 1D CNN approach will be helpful for real-time monitoring of activities and for tracing the behavior of dogs.

This article presents the first keyword spotting (KWS) IC that uses a ring-oscillator-based time-domain processing technique for its analog feature extractor (FEx). Its extensive usage of time-encoding schemes allows the analog audio signal to be processed in a fully time-domain manner except for the voltage-to-time conversion stage of the analog front end. Benefiting from fundamental building blocks based on digital logic gates, it offers better technology scalability compared to conventional voltage-domain designs. Fabricated in a 65-nm CMOS process, the prototyped KWS IC occupies 2.03 mm 2 and dissipates 23- $\mu \text{W}$ power consumption, including analog FEx and digital neural network classifier. The 16-channel time-domain FEx achieves a 54.89-dB dynamic range for 16-ms frame shift size while consuming 9.3 $\mu \text{W}$. The measurement result verifies that the proposed IC performs a 12-class KWS task on the Google Speech Command dataset (GSCD) with >86% accuracy and 12.4-ms latency. [ABSTRACT FROM AUTHOR]

Network security requires effective detection and proper analysis of abnormal network behavior. To address the uncertainty associated with the process of network intrusion detection, this article proposes a network intrusion-detection algorithm based on dynamic intuitionistic fuzzy sets (IFSs). We use the classic network intrusion datasets KDD 99, NSL-KDD, and the massive, high-dimensional dataset UNSW-NB15 to evaluate the performance of our proposed algorithm. First, we perform data preprocessing on these three datasets and select features based on the results of a chi-square test. Second, using time-series processing, we construct dynamic intuitionistic fuzzy patterns from the feature-selected datasets. At last, we use a proposed distance measure for the dynamic IFSs to generate a classifier that facilitates the detection of network intrusion. Experimental results show that the classification performance of the proposed algorithm is superior to that of other state-of-the-art algorithms on the three aforementioned datasets. The achieved improvement in classification performance is particularly significant for large datasets. [ABSTRACT FROM AUTHOR]

Diabetic sensorimotor polyneuropathy (DSPN) is an early indicator for non-healing diabetic wounds and diabetic foot ulcers, which account for one of the most common complications of diabetes, leading to increased healthcare cost, decreased quality of life, infections, amputations, and death. Early detection and intelligent classification tools for DSPN can allow correct diagnosis and treatment of painful diabetic neuropathy as well as a timely intervention to prevent foot ulceration, amputation, and other diabetic complications. Hence, to successfully mitigate the prevalence of DSPN, this study aims to depict an intelligent DSPN severity classifier using Adaptive Neuro Fuzzy Inference System (ANFIS). Michigan Neuropathy Screening Instrumentation (MNSI) was considered as the input for identification and stratification of DSPN. Patients have been classified into four classes: Absent, Mild, Moderate, and Severe. The model accuracy was validated with the results from different machine learning algorithms. The Accuracy, sensitivity, and specificity of the ANFIS model are 91.17±1.18%, 92±2.26%, 96.72±0.93%, respectively. The proposed classifier was used to classify the Epidemiology of Diabetes Interventions and Complications (EDIC) clinical trial patients and observed that in the first, eighth, and nineteenth EDIC years 18.31%, 39.45%, and 59.14% patients had different levels of DSPN. This study also investigates the changes in muscle activity during gait from three different lower limb muscles (vastus lateralis (VL), tibialis anterior (TA), and gastrocnemius medialis (GM)) electromyography (EMG) of DSPN patients with different severity levels classified by the proposed classifier and observed that VL and GM muscles show an increase in delay for activation peak and decrease in peak magnitude during gait with the progression of DSPN severity. Based on this observation, the ANFIS model was trained using the extracted EMG features for DSPN severity stratification and showed promising results. Our proposed ANFIS based severity classifier using both MNSI variables and EMG features will help health professionals to diagnose and stratify DSPN severity based on both signs and symptoms and electrophysiological changes due to DSPN.

Sentiment analysis is crucial for studying public opinion since it can provide us with valuable information. Existing sentiment analysis methods rely on finding the sentiment element from the content of user-generated. However, the question of why a message produces certain emotions has not been well explored or utilized in previous works. To address this challenge, we propose a natural language explanation framework for sentiment analysis that provides sufficient domain knowledge for generating additional labelled data for each new labelling decision. A rule-based semantic parser transforms these explanations into programmatic labelling functions that generate noisy labels for an arbitrary amount of unlabelled sentiment information to train a sentiment analysis classifier. Experiments on two sentiment analysis datasets demonstrate the superiority it achieves over baseline methods by leveraging explanations as external knowledge to joint training a sentiment analysis model rather than only labels. An ablation study is conducted to clarify the relative contribution of natural language explanations.

During the last decade, Speech Emotion Recognition (SER) has emerged as an integral component within Human-computer Interaction (HCI) and other high-end speech processing systems. Generally, an SER system targets the speaker’s existence of varied emotions by extracting and classifying the prominent features from a preprocessed speech signal. However, the way humans and machines recognize and correlate emotional aspects of speech signals are quite contrasting quantitatively and qualitatively, which present enormous difficulties in blending knowledge from interdisciplinary fields, particularly speech emotion recognition, applied psychology, and human-computer interface. The paper carefully identifies and synthesizes recent relevant literature related to the SER systems’ varied design components/methodologies, thereby providing readers with a state-of-the-art understanding of the hot research topic. Furthermore, while scrutinizing the current state of understanding on SER systems, the research gap’s prominence has been sketched out for consideration and analysis by other related researchers, institutions, and regulatory bodies.

One common computer vision task is to track an object as it moves from frame to frame within a video sequence. There are a myriad of applications for such capability and the underlying technologies to achieve this tracking are very well understood. More recently, deep convolutional neural networks have been employed to not only track, but also to classify objects as they are tracked from frame to frame. These models can be used in a tracking paradigm known as tracking by detection and can achieve very high tracking accuracy. The major drawback to these deep neural networks is the large amount of mathematical operations that must be performed for each inference which negatively impacts the number of tracked frames per second. For edge applications residing on size, weight, and power limited platforms, such as unmanned aerial vehicles, high frame rate and low latency real time tracking can be an elusive target. To overcome the limited power and computational resources of an edge compute device, various optimizations have been performed to trade off tracking speed, accuracy, power, and latency. Previous works on motion based interpolation with neural networks either do not take into account the latency accrued from camera image capture to tracking result or they compensate for this latency but are bottlenecked by the motion interpolation operation instead. The algorithm presented in this work gains the performance speedup used in previous motion based neural network inference papers and also performs a novel look back operation that is less cumbersome than other competing motion interpolation methods.

Traditional Image Processing Techniques (IPT), used for automating the detection and classification of weld defects from radiography images, have their own limitations, which can be overcome by Deep Neural Networks (DNN). DNN produces considerably good results in fields which offer big dataset for it to train. DNN trained with small datasets by conventional methods produces less accurate results. This limits the use of DNN in many fields. This study focuses to overcome this limitation, by adopting transfer learning using Pre-trained deep convolutional neural networks. By this method, a weld defect radiographic image classifier, which can classify 14 types of weld defects, was constructed. 940 Image patches of weld defects were manually collected and labelled from GDXray database. Subsequently the features of this weld defect dataset were extracted using VGG16 and ResNet50 CNNs, both pre-trained on the ImageNet database. Then machine learning models such as Logistic Regression, Support Vector Machine (SVM) and Random Forest were trained on these extracted features. The Classifier based on SVM trained on features extracted by ResNet50 outperforms the other counter parts with an accuracy of 98%. In all these cases, transfer learning improves performance and reduces the training time and computational system requirements.

The most important parts of rotating machinery are the rolling bearings. Finding bearing faults in time can avoid affecting the operation of the entire equipment. The data-driven fault diagnosis technology of bearings has recently become a research hotspot, and the starting point of research is often the acquisition of vibration signals. There are many public data sets for rolling bearings. Among them, the most widely used public dataset is Case Western Reserve University bearing center (CWRU). This paper will start from the CWRU data set, compare and analyze some basic methods of machine learning based rolling bearing fault diagnosis, and summarize the characteristics of CWRU. First, we give a comprehensive introduction to CWRU and summarize the results achieved. After that, the basic methods and principles of machine learning based rolling bearing fault diagnosis were summarized. Finally, we conduct experiments and analyze experimental results. This paper will have certain guiding significance for the future use of CWRU for machine learning based rolling bearing fault diagnosis.

Indoor scene understanding is an active research topic that has recently gained increasing attention in both computer graphics and computer vision community. This paper presents a novel automatic point cloud instance abstraction and segmentation of indoor scenes using learned pairwise planar patch relations. Planar patches have sufficient representative power to abstract the semantics in mostly man-made 3D indoor scenes. To exploit the planar patch relationships, a new indoor point clouds dataset is automatically derived from the public dataset ScanNet, which can be adopted for training a pairwise patch relation classifier. Given the pre-processed indoor scene point clouds, a set of planar patches are generated to provide robust and compact representation of indoor objects and parts. For each extracted planar patch, a feature descriptor in view of its geometric properties and appearance is calculated. Owing to the proposed feature descriptors, the relations of pairwise planar patch are measured and can be fed into the pre-trained patch relation classifier. To yield the instance segmentation of indoor scenes, a robust graph-based algorithm is adopted to group these planar patches which incorporate the learned patch relations. Experimental results demonstrate that our proposed approach can produce instance semantic segmentation of various cluttered indoor scenes robustly and efficiently.

This paper presents the concept of Graph-based Local Resampling of perceptron-like neural networks with random projections (RN-ELM) which aims at regularization of the yielded model. The addition of synthetic noise to the learning set finds some similarity with data augmentation approaches that are currently adopted in many deep learning strategies. With the graph-based approach, however, it is possible to direct resample in the margin region instead of exhaustively cover the whole input space. The goal is to train neural networks with added noise in the margin region, located by structural information extracted from a planar graph. The so-called structural vectors, which are the training set vertices near the class boundary, are obtained from the structural information using Gabriel Graph. Synthetic samples are added to the learning set around the geometric vectors, improving generalization performance. A mathematical formulation that shows that the addition of synthetic samples has the same effect as the Tikhonov regularization is presented. Friedman and pos-hoc Nemenyi tests indicate that outcomes from the proposed method are statistically equivalent to the ones obtained by objective-function regularization, implying that both methods yield smoother solutions, reducing the effects of overfitting.

To address the problem that the traditional human motion attitude detection process is easy to ignore the data calibration, which leads to the problems of long running time, low accuracy and poor detection effect, a human motion target attitude detection algorithm based on semi-supervised learning in the Internet of things environment is proposed. Firstly, human motion target images are collected using the Internet of things (IoT), human motion attitude features are extracted based on the eight-star model, and multi-features are fused to form image blocks of 17-dimensional feature vectors. Then, random fern classifiers are optimized and semi-supervised learning is used to calculate a large number of uncalibrated data in time domain, spatial domain and data. The classifier is trained to complete image block classification. Finally, the classifier parameters are updated iteratively to complete the attitude detection of human motion target. The results show that the proposed algorithm has high accuracy in human motion attitude extraction and multi-feature fusion, and has a high correct classification rate for different feature poses, as high as 92.5%. The average F value of human motion attitude detection is 0.95, the overlap ratio is high and the time is short. The overall performance is good.

A precise knowledge about future traffic will eventually open a new era in traffic management. Research has focused on the still unresolved problem of predicting travel time (TT). However, practitioners favor the level of service (LOS) as a meaningful metric that avoids the continuous fluctuations and link-specificity of TT. Evolving from TT to LOS opens a new research line in the field, moving the underlying mathematical problem from regression to classification. This study proposes a short-term LOS classifier to fulfill this requirement. Given that traffic conditions are mostly free-flow throughout the day, LOS classes are unbalanced. Therefore, we based our predictor on a Random Undersampling Boost algorithm (RUSBoost), especially suited to overcome this issue. We trained and validated this LOS predictor with 12 months of arrival travel time data, captured by a Bluetooth network with 6 links, in real operation on the SE-30 highway (Seville, Spain). This classifier achieved an average recall of 82.8% for prediction horizons up to 15 minutes, reaching 92.5% predicting congestion. We reached this performance by exploiting two facts that we empirically demonstrated: (i) information from every link (even those in the opposite direction) contributes to increase the accuracy of the prediction; and (ii) traffic presents different behavior depending on the day of the week, which we used to segment the data and construct specific classifiers. These promising results show the potential of the proposed LOS predictor, providing a new perspective into traffic forecast and the subsequent traffic management that yields with what practitioners demand. [ABSTRACT FROM AUTHOR]

At present, regions of the same class determined by Support Vector Machines (SVM) classifier, Support Vector Domain Description (SVDD) classifier and Deep Learning (DL) classifier may occupy regions of other classes or unknown classes in feature space. There exists a risk that samples of other classes or unknown classes are wrongly classified as a known class. In this paper, the Support Vector Domain Tightly Wrapping Description Design (SVDTWDD) method with appropriate rejection regions and the corresponding incremental learning algorithm are proposed to overcome the above problem. The main work includes: (1) We develop a construction algorithm of the tightly wrapping set for the homogeneous feature set; (2) Based on the homogeneous feature set and tightly wrapping set, a novel algorithm is presented for obtaining the tightly wrapping surface of the homogeneous feature region; (3) The method for constructing all the public regions outside of the tightly wrapping surface and the intersections of wrapping regions in two different tightly wrapping surfaces, as the rejection region of the classifier; (4) An incremental algorithm is also presented based on the SVD-TWDD method. The experimental results with UCI data sets show that the correct recognition rate of our proposed method is nearly100% even if with a low rejection rate.

Emotions can be evoked in humans by images. Previous reports on Recognition of Emotions induced by Visual Content of images (REVC) mainly focused on numerous features to improve recognition performance. To devise a more robust REVC system, this paper examines the performance of a wide range of classifiers using color histogram as a single feature. Different numbers of color histogram bins in both RGB (red, green, blue) and HSV (hue, saturation, value) color spaces are considered in the examination and the overall classification performance is compared across the bin sizes. This investigation shows that features are not the only important factors affecting the performance of REVC systems, but also the type of classifiers and their parameters. This study shows that the HSV color space is better suited than the RGB color space for REVC systems. This paper proposes a new optimization algorithm called Optimizing Parameters of Ensemble RUSboosted Tree (OPERT) to boost the performance of the REVC system. Furthermore, a novel REVC system called Color histogram with Optimized RUSboosted Tree (CORT) is introduced. It is shown that our method is simpler, faster, and more efficient than the state-of-the-art, while providing comparable recognition performance. The robustness of the CORT system is validated over three different image datasets.

Non-intrusive load monitoring (NILM) discerns the individual electrical appliances of a residential or commercial building by disaggregating the accumulated energy consumption data without accessing to the individual components applying a single-point sensor. The fundamental concept is to decompose the aggregate load into a family of appliances that can explain its characteristics. In the age of smart grid networks and sophisticated energy management infrastructures, NILM can be considered as a significant tool pertaining to smart and inexpensive energy metering technique. In this article, a novel NILM solution based on capsule network is proposed, where convolutional neural network (CNN) is employed to extract potential features from a set of non-overlapping energy measurement data segments and the capsule architecture is designed to predict class probabilities of the individual segments. Then, a decision making algorithm is proposed to compute the final classification based on the predicted class probabilities of the segments. The presented research design comprises two unique NILM applications - device type classification from individual sensor recordings stored in COOLL and PLAID public databases, and device activity status monitoring at any particular time instant from aggregated energy consumption data recorded in UK-DALE database. Additionally, substantial experimental investigations have been carried out for device type classification accounting on various types of train and test set distributions as well as individual instrument and house classifications. The presented framework analyzes different parameters and metrics in depth to corroborate the efficacious performance evaluations for real-time applications. Relevant performance comparisons with existing works in literature validate the sustainability of the proposed solution.

Sensor-cloud infrastructure provides a storage platform for the massive sensed data, that is flexible and re-configurable, for various application areas which are monitored through the resource-limited networks such as wireless sensor networks (WSNs), ad hoc networks, and Internet of things (IoT). Due to their overwhelming characteristics, these networks are used in different application areas to assist human beings in their daily-life activities. However, these networks have different challenging issues such as reliability in communication and processing, storage of the massive data, efficient utilization of on-board battery, maximum lifetime achievement, minimum possible average packet loss ratio, and reliable routing mechanisms. Although various communication and load balancing mechanisms have been proposed in the literature to resolve this issue, however, these schemes are either application specific or overlay complex. In this paper, a reliable communication and load balancing scheme for the resource-limited networks is presented to resolve these issues, particularly with available resources. To achieve these goals, the proposed scheme bounds every sensing device Ci to compute the transmission capabilities of its neighboring devices that is residual energy Er, hop count Hc, round trip time (RTTi), and processing cost. Initially, to guarantee reliable wireless communication, a source device prefers a neighboring device Ci with minimum Hc value over those having maximum Hc values. Moreover, this scheme bounds every device Ci to find four shortest & reliable paths and forward maximum packets on two of these paths preferably on the most reliable and optimal route. Therefore, unlike the traditional shortest path scheme, devices Ci reside on these paths do not deplete their on-board battery more rapidly than others. To further improve the reliability of the proposed scheme, the assigned weight-age factors are fine-tuned if one or two of the neighboring devices Ci consume 80% of their on-board battery, that is now maximum weight-age is assigned to the residual energy Er and minimum to Hc value respectively. Simulation results show the exceptional performance of the proposed reliable communication and load balancing scheme against the field-proven schemes in terms of average packet delivery ratio, average throughput, end-to-end delay, and overall network lifetime.

The main work of human motion gesture recognition is to recognize and analyze the behavior of human objects in the video. Although the current research in the field of human motion gesture recognition has achieved certain results, the human motion gesture recognition in real life scenes has great effects due to factors such as camera movement, target scale transformation, dynamic background, viewing angle, and illumination. This article first proposes a new method of constructing human motion posture features to describe human behavior. This method is based on deep convolutional neural network features and topic models. Experiments have verified that compared with the traditional feature map extracted from the convolutional neural network fully connected layer, the feature map extracted from the convolutional neural network convolutional layer is not only lower in dimension but also has higher discrimination. Secondly, based on the feature map of the convolutional neural network, the training map downsampling strategy is used to overcome the interference caused by the object's scale change and shape change. Finally, based on the basketball gesture recognition method, the behavior performance of the legs and arms in 9 basketball actions of walking, running, jumping, standing dribbling, walking dribbling, running dribbling, shooting, passing and receiving is analyzed. As well as the corresponding signal waveform characteristics, a two-stage data division method for basketball is proposed. The unit action data is extracted for analysis to realize feature extraction. In order to select the most suitable classifier for basketball gesture recognition, the constructed feature vector uses four Different classifiers are trained to construct different classifiers to realize the division of actions.

Medical research confirms that eye movement abnormalities are related to a variety of psychological activities, mental disorders and physical diseases. However, as the specific manifestations of various diseases in terms of eye movement disorders remain unclear, the accurate diagnosis of diseases according to eye movement is difficult. In this paper, a deep neural network (DNN) method is employed to establish a disease discrimination model according to eye movement. First, multiple eye-tracking experiments are designed to obtain eye images. Second, pupil characteristics, including position and size, are extracted, and the feature vectors of eye movement are obtained from the normalized pupil information. Based on a long short-term memory (LSTM) network, a classifier that corresponds to each feature, which is referred to as a weak classifier, is built. The experimental samples are preclassified, and the classification ability of each weak classifier for different diseases is also calculated. Last, a strong classifier is achieved for disease discrimination by synthesizing all the weak classifiers and their classification abilities. By classification testing for three categories of healthy controls, brain injury patients and vertigo patients, the experimental results demonstrated the efficiency of this method. With the deep learning method, more medical information can be excavated from eye movement to improve the values in disease diagnosis.

This study proposed a wearable device capable of recognizing six human daily activities (walking, walking upstairs, walking downstairs, sitting, standing, and lying) through a deep learning algorithm. Existing wearable devices are mainly watches or wristbands, and almost none are to be worn on the waist. Wearable devices in the forms of watches and wristbands are unfriendly to patients who are critically ill, such as patients undergoing dialysis. Patients undergoing dialysis have artificial blood vessels on their arm, and they cannot perform intense exercise. For this type of users, general hand wearable devices cannot correctly identify wearers' activities. Therefore, we proposed a waist wearable device and these types of daily life activities to assess their exercise. The hardware of the wearable device consisted of an inertial sensor, which included a microcontroller, a three-axis accelerometer, and a three-axis gyroscope. The activity recognition algorithm of the software used motion signals acquisition, signal normalization, and a feature learning method. The feature learning method was based on a 1D convolutional neural network that automatically performed feature extraction and classification from raw data. One part of the experimental data was from the dataset of the University of California (UCI), and the other part was recorded by this study. To capture the data recorded, the wearable inertial sensing device was attached to the waists of 21 experimental participants who performed six common movements in a laboratorial environment, and the subsequent records were collected to verify the validity of the proposed deep learning algorithm in relation to the inertial sensor of the wearable device. For the six common activities in the UCI dataset and the data recorded, the recognition rates in the training sample reached 98.93% and 97.19%, respectively, and the recognition rates in the testing sample were 95.99% and 93.77%, respectively.

Around the world, fossil fuels are decreasing and pollution is increasing. As a new energy source, microbial fuel cells (MFCs) have been widely concerned. However, most of the previous researches focused on the material selection, configuration design and optimal control of MFCs, and few of them were able to systematically analyze the failures of MFCs. In order to ensure the reliable operation of MFCs, this paper systematically explores the MFC fault diagnosis process, including the acquisition of initial fault data, feature extraction and fault classification.Firstly, in order to acquire data quickly and effectively, the mathematical model is used to simulate the occurrence of faults, and four types of typical fault voltages are obtained. Then, wavelet analysis is used to extract the voltage characteristics of MFC faults, and the characteristics of each fault are explored in eight frequency bands. Finally, the recognition effects of various classifiers on fault features are compared. Through the analysis of the results, it is found that fault tree is the most suitable fault diagnosis method for MFCs. The fault data extraction method proposed in this paper and the classification effect of various classifiers finally obtained provide a reference for the further analysis of MFC faults. At the same time, the combination of wavelet analysis and fault tree diagnosis model proposed in this paper provides ideas for fault diagnosis in other fields.

Polymer modified Quartz Tuning Fork sensor array with three sensors has been developed to classify 1,4-dimethoxy-2, 3-butanediol (BD), cyclohexanone (CH), ethanol, methanol and acetone with various concentrations. Linear Discriminant Analysis (LDA), Decision Tree (DT), and Random Forrest (RF) have been used and evaluated as the classifiers predicting the VOCs and their concentration. The classifiers were trained with three features: response time, recovery time, and shift in the resonant frequency of the sensors. The results have been compared and analyzed for the performance of each classifier for the present system. RF gives the best performance as a classifier with an accuracy of classification more than 95%. We present results of VOC classifier performance for the QTF sensor array. [ABSTRACT FROM AUTHOR]

As a key component of the high-speed railway catenary, the insulator plays an important role in supporting the catenary and maintaining the insulation between the catenary and earth. Regular inspection using computer vision technology is an effective way to detect insulator defects and improve the catenary operation safety. However, achieving full automation of insulator defects detection is still a difficult task due to the defective sample scarcity and subtle defect features. To overcome the problem, this article proposes defect detection GANs (DefGANs) that consist of a denoising autoencoder, a discriminator, and a classifier to detect defects, which can generate defective samples from pitted latent representations, thereby improving the reliability of defect detection classifier. The proposed DefGAN can pit the latent representations of normal samples to generate defective samples. Our proposal mainly consists of two stages. In the first stage, we use cascaded deep segmentation networks (CDSNets) to extract the insulator from the catenary images. Then, the insulator area is sampled into small patches. In the second stage, the patches are fed into the proposed DefGAN to detect defects. Moreover, the defect score is determined by the anomaly probability of the classifier and reconstruction error of the denoising autoencoder. The effectiveness of our work is measured across the catenary data sets. [ABSTRACT FROM AUTHOR]

It is well known that the social network users' behavior can reflect their personal mental profiles. Therefore, in this paper, we conducted an experiment to study users' gender identity through a mobile social network. Gender identity is the personal sense of one's own gender, which can be the same or different from their sex assigned at birth, that is, one's innermost concept of self as male, female, a blend of both, or neither. In general, a person's psychological gender cannot be observed from the external characteristics of people. And even if some user's attributes such as physical gender and age can be found in social networks, it is difficult to find out the psychological gender of users. Apart from the fact that it will be hidden as the user's privacy, the main reason is that users' behaviors in social networks have no direct relationship with psychological gender. In this paper, we choose a special mobile social network to get around these limitations. Specifically, this paper is based on a real-world large gay social network of more than 10 000 000 users. In this mobile social network, we can indirectly mark the psychological gender of users by their sexual orientation. We collect 1000 users, including their characters, followers, followees and the same data for their followers and followees. Then, we examine an extensive set of features that can be extracted from the data set and train some classifiers to automatically identify the user's psychological gender. In the special mobile social network, a user's psychological gender is related to his characteristics, which include his personal characteristics, dating preference, and so on. This paper mainly analyzes users' personal characteristics, triads, and triadic evolution to infer their psychological gender. The experiments show that all these features are indeed effective in the classification, and even the features considered in previous studies have different implications with gender identity on other social networks. And it turns out that psychological gender as a kind of privacy can be leaked by these unintentional behaviors, and so we give some practical suggestions at the end of this paper to protect users' privacy.

Due to the rapid development of Internet technologies and social media, sentiment analysis has become an important opinion mining technique. Recent research work has described the effectiveness of different sentiment classification techniques ranging from simple rule-based and lexicon-based approaches to more complex machine learning algorithms. While lexicon-based approaches have suffered from the lack of dictionaries and labeled data, machine learning approaches have fallen short in terms of accuracy. This paper proposes an integrated framework which bridges the gap between lexicon-based and machine learning approaches to achieve better accuracy and scalability. To solve the scalability issue that arises as the feature-set grows, a novel genetic algorithm (GA)-based feature reduction technique is proposed. By using this hybrid approach, we are able to reduce the feature-set size by up to 42% without compromising the accuracy. The comparison of our feature reduction technique with more widely used principal component analysis (PCA) and latent semantic analysis (LSA) based feature reduction techniques have shown up to 15.4% increased accuracy over PCA and up to 40.2% increased accuracy over LSA. Furthermore, we also evaluate our sentiment analysis framework on other metrics including precision, recall, F-measure, and feature size. In order to demonstrate the efficacy of GA-based designs, we also propose a novel cross-disciplinary area of geopolitics as a case study application for our sentiment analysis framework. The experiment results have shown to accurately measure public sentiments and views regarding various topics such as terrorism, global conflicts, and social issues. We envisage the applicability of our proposed work in various areas including security and surveillance, law-and-order, and public administration.

Gait phase detection is an important procedure in the application of many lower limb auxiliary robots. The gait phase detection algorithms that utilize surface electromyography (sEMG) signals have been developed to overcome unnatural walking. However, traditional studies mostly consider precise gait walking events when the subject is focusing only on walking, and the accuracy of traditional algorithms is susceptible to more realistic gait scenarios, such as walking with cognitive tasks. In this paper, the gait phase detection is considered in more realistic and challenging scenarios, in which the subject is walking while performing cognitive tasks. A kernel linear discriminant analysis (LDA)-based nonlinear fusion model is proposed for gait recognition, which can effectively reduce the error caused by cognitive tasks, making it an ideal model for gait phase detection while cognitive tasks are performed in the process of walking. Furthermore, the Internet of Things (IoT) framework is incorporated to reduce the gait phase detection algorithm's process time by offloading the data from local sEMG sensors to the IoT server with powerful computation capability. The experiments have been conducted to validate our proposed algorithms, demonstrating that the boundaries between the stance period and swing period are more blurred when walking with cognitive tasks.

Inverter air conditioners (IACs) with considerable total capacity and fast response speed are ideal demand response resources, which are of significant potential to provide reserve capacity for the power system frequency regulation. However, due to the complexity and implicitness of the frequency response models, it is difficult to formulate the optimization problem considering frequency dynamics to allocate reserve capacity precisely. In this paper, a data-driven method is proposed for reserve allocation with the frequency security constraint considering IACs. Firstly, the equivalent frequency response model of aggregated IACs is developed considering electrical-thermal characteristics and then incorporated into the frequency regulation framework of power systems along with conventional generators. Then, simulations are implemented to generate massive reserve samples with deterministic frequency security labels. Later, a support vector machine (SVM) based frequency security classifier is trained to convert the implicit frequency security constraint into polynomials and reshape the reserve allocation problem into a solvable general quadratically constrained quadratic program (QCQP). Finally, a heuristic Suggest-and-Improve (SI) method is adopted to deal with the nonconvex QCQP of interest. It is demonstrated by numerical studies that the proposed data-driven method enables power systems to operate closer to the frequency security boundaries and thus achieve lower costs.

Cascading failure may incur catastrophic consequences. The high order and non-simultaneousness properties are the leading causes for the massive power flow (PF) calculations of the online static security assessment (SSA). Current PF models fail to satisfy the requirements for accuracy, speed, and robustness. This paper proposes a novel method where line outage distribution factor (LODF) model and AC model are coordinated by a binary classifier. The classifier determines the feasibility of LODF model in a specific case by evaluating its potential error. Cases are preferred to be analyzed by LODF model, while the ones with large potential errors will be analyzed by AC model. Case studies of three IEEE systems and the Texas 2000-bus system are given to verify the effectiveness, flexibility, and robustness of the proposed method.

Cancer is one of the most influential factors causing death in the world. Adenosine which is a molecule, found in all human cells by coupling with G protein it turns into an adenosine receptor. Adenosine receptor is an important target for cancer therapy. Adenosine stops the growth of malignant tumor cells such as lymphoma, melanoma and prostate carcinoma. Adenosine is activated by interacting with drugs to stop tumor cells from spreading and cure cancer disease. This research aims to predict drugs and potential drug candidates that interact with adenosine receptors. We built a machine learning model using three different classification techniques: Random Forest (RF), Decision Tree (DT) and Support Vector Machine (SVM) then we chose the best technique after comparing the results. Unlike other researches, we used the drug side effect integrated into drug fingerprint as a feature to train our model to classify drugs (interacting and non-interacting) with adenosine receptors. We ranked the interacting drugs with adenosine receptors based on drug side effects to find the most preferred drug (least side effect) among several drugs, which helps in drug design. Most existing datasets contain drugs, targets and the interactions between them, neglecting drug side effects. We formed a new dataset that has the drug side effect. The new dataset is composed of 400 drugs, 794 targets and 3990 drug side effects. Since the dataset was imbalanced we applied Synthetic Minority Oversampling Technique (SMOTE). After conducting experiments, RF achieved the best classification performance with an accuracy of 75.09%.

Autism Spectrum Disorder (ASD) is a group of neurodevelopmental disabilities that are not curable but may be ameliorated by early interventions. We gathered early-detected ASD datasets relating to toddlers, children, adolescents and adults, and applied several feature transformation methods, including log, Z-score and sine functions to these datasets. Various classification techniques were then implemented with these transformed ASD datasets and assessed for their performance. We found SVM showed the best performance for the toddler dataset, while Adaboost gave the best results for the children dataset, Glmboost for the adolescent and Adaboost for the adult datasets. The feature transformations resulting in the best classifications was sine function for toddler and Z-score for children and adolescent datasets. After these analyses, several feature selection techniques were used with these Z-score-transformed datasets to identify the significant ASD risk factors for the toddler, child, adolescent and adult subjects. The results of these analytical approaches indicate that, when appropriately optimised, machine learning methods can provide good predictions of ASD status. This suggests that it may possible to apply these models for the detection of ASD in its early stages.

This paper illustrates the most recent results regarding the criticality ranking decision support classifier for an individual business function. The validated classifier is part of the business continuity points approach which estimates the recovery complexity of an individual business function and is based on the use case points method for estimating the software complexity. The business continuity points method is utilized in order to estimate specific recovery complexity parameters of a given business function. A part of the approach concerns the business function criticality ranking which is based on the recovery complexity parameters. In this paper, we measure the accuracy in the criticality ranking classifier by comparing results between the speedy and the detailed criticality ranking of a business function. The measurement is performed via the R-Studio software and the confusion matrix technique. The results are based on a learning data set prepared in MS Excel which includes the empirical calculations for constructing the specific classifier.

The paper discusses a new framework combining the possibilities of Big Data processing and machine leaning developed for security monitoring of mobile Internet of Things. The mathematical foundations and the problem statement are considered. The description of the used data set and the architecture of proposed security monitoring framework are provided. The framework specifies several machine learning mechanisms intended for solving classification tasks. The classifier operation results are exposed to plurality voting, weighted voting, and soft voting. The framework performance and accuracy is assessed experimentally.

Currently, carbon dioxide (CO2) waveforms measured by capnography are used to estimate respiratory rate and end-tidal CO2 (EtCO2) in the clinic. However, the shape of the CO2 signal carries significant diagnostic information about the asthmatic condition. Previous studies have shown a strong correlation between various features that quantitatively characterize the shape of CO2 signal and are used to discriminate asthma from non-asthma using pulmonary function tests, but no reliable progress was made, and no translation into clinical practice has been achieved. Therefore, this paper reports a relatively simple signal processing algorithm for automatic differentiation of asthma and non-asthma. CO2 signals were recorded from 30 non-asthmatic and 43 asthmatic patients. Each breath cycle was decomposed into subcycles, and features were computationally extracted. Thereafter, feature selection was performed using the area (Az) under the receiver operating characteristics curve analysis. A classification was performed via a leave-oneout cross-validation procedure by employing a support vector machine. Our results show maximum screening capabilities for upward expiration (AR1), downward inspiration (AR2), and the sum of AR1 and AR2, with an Az of 0.892, 0.803, and 0.793, respectively. The proposed method obtained an average accuracy of 94.52%, sensitivity of 97.67%, and specificity of 90% for discrimination of asthma and non-asthma. The proposed method allows for automatic classification of asthma and non-asthma condition by analyzing the shape of the CO2 waveform. The developed method may possibly be incorporated in real-time for assessment and management of the asthmatic conditions.

Prediction of protein subcellular location has currently become a hot topic because it has been proven to be useful for understanding both the disease mechanisms and novel drug design. With the rapid development of automated microscopic imaging technology in recent years, classification methods of bioimage-based protein subcellular location have attracted considerable attention for images can describe the protein distribution intuitively and in detail. In the current study, a prediction method of protein subcellular location was proposed based on multi-view image features that are extracted from three different views, including the four texture features of the original image, the global and local features of the protein extracted from the protein channel images after color segmentation, and the global features of DNA extracted from the DNA channel image. Finally, the extracted features were combined together to improve the performance of subcellular localization prediction. From the performance comparison of different combination features under the same classifier, the best ensemble features could be obtained. In this work, a classifier based on Stacked Auto-encoders and the random forest was also put forward. To improve the prediction results, the deep network was combined with the traditional statistical classification methods. Stringent cross-validation and independent validation tests on the benchmark dataset demonstrated the efficacy of the proposed method. [ABSTRACT FROM AUTHOR]

This work has as main objective the development of a soft-sensor to classify, in real time, the behaviors of drivers when they are at the controls of a vehicle. Efficient classification of drivers’ behavior while driving, using only the measurements of the sensors already incorporated in the vehicles and without the need to add extra hardware (smart phones, cameras, etc.), is a challenge. The main advantage of using only the data center signals of modern vehicles is economical. The classification of the driving behavior and the warning to the driver of dangerous behaviors without the need to add extra hardware (and their software) to the vehicle, would allow the direct integration of these classifiers into the current vehicles without incurring a greater cost in the manufacture of the vehicles and therefore be an added value. In this work, the classification is obtained based only on speed, acceleration and inertial measurements which are already present in many modern vehicles. The proposed algorithm is based on a structure made by several Neurofuzzy systems with the combination of projected data in components of various Principal Component Analysis. A comparison with several types of classical classifying algorithms has been made. [ABSTRACT FROM AUTHOR]

In this article, the accuracy of inverter-based memristive neural networks (NNs) for function approximation applications is improved under the presence of process variations. The improvement is achieved by using a design approach, called INTERSTICE (Inverter-based Memristive Neural Networks Dis cretization for Function Approximation Applications), which discretizes the output values by employing a classifier. More precisely, in the INTERSTICE approach, the output range is divided into $K$ subranges where each subrange is considered as a class. To train the classifier, the training samples are labeled where each label shows belonging to a specific class. To evaluate the efficacy of the design technique, some function approximation applications such as BlackScholes, FFT, $K$ -means, and Sobel are considered. Compared to PHAX, a recently published inverter-based memristive NN, INTERSTICE provides lower mean squared error (MSE) values in the presence of memristor and transistor variations. More specifically, the improvements in the mean of MSE ($\mu _{\mathrm {MSE}}$) are in the range of 40%–80% when considering 10% variations in the memristor resistance and transistor parameters. In addition, for most of the benchmarks, INTERSTICE improves the $\mu _{\mathrm {MSE}}$ values of the nominal case (the case where all circuit elements are ideal) compared to PHAX. As another advantage compared to the PHAX, in INTERSTICE, digital outputs can be generated based on the selected classes which eliminates the need for an analog-to-digital converter at the output port connected to the digital part of the system. Finally, achieving lower $\mu _{\mathrm {MSE}}$ values using fewer memristors and consuming lower energy is also attainable with this design approach. [ABSTRACT FROM AUTHOR]

The quantity and quality of training samples have a great influence on the performance of most hyperspectral image classification approaches. However, in a real scenario, manually annotating a large number of accurate training samples is extremely labor-intensive and time-consuming. In this article, a multilabel training sample augmentation method is proposed. Instead of giving an exact label to each pixel, we just precisely label a small number of pixels by giving them a single label (called single-label samples) and annotate a large number of pixels in certain regions together by giving them multiple labels (called multilabel samples). Furthermore, in order to make full use of the multilabel training samples, a superpixel segmentation and recursive filtering-based method is proposed. The proposed method consists of the following major steps: recursive filtering-based feature extraction, superpixel-based segmentation, and spectral–spatial similarity-based mislabeled sample removal. Experimental results demonstrate that the proposed method can significantly improve the classification accuracy of multiple classifiers by using the multilabel training samples. [ABSTRACT FROM AUTHOR]

In long-term electrocardiogram (ECG) recording for arrhythmia monitoring, using a uniform compression strategy throughout the entire data to achieve high compression efficiency may result in unacceptable distortion of abnormal beats. The presented work addressed a solution to this problem, rarely discussed in published research. A support vector machine (SVM)-based binary classifier was implemented to identify the abnormal beats, achieving a classifier sensitivity (SE) and negative predictive value (NPV) of 99.89% and 0.003%, respectively with 34 records from MIT-BIH Arrhythmia database (mitdb). A hybrid lossy compression technique was implemented to ensure on-demand quality, either in terms of distortion or compression ratio (CR) of ECG data. A wavelet-based compression for the abnormal beats was implemented, while the consecutive normal beats were compressed in groups using a hybrid encoder, employing a combination of wavelet and principal component analysis. Finally, a neural network-based intelligent model was used, which was offline tuned by a particle swarm optimization (PSO) technique, to allocate optimal quantization level of transform domain coefficients generated from the hybrid encoder. The proposed technique was evaluated with four types of morphology tags, “A,” “F,” “L,” and “V,” from mitdb database, achieving less than 2% PRDN and less than 1% in two diagnostic distortion measures for abnormal beats. Overall, an average CR of 19.78 and PRDN of 3.34% was obtained. A useful outcome of the proposed technique is the low reconstruction time in rapid screening of long arrhythmia records, while only abnormal beats are presented for evaluation. [ABSTRACT FROM AUTHOR]

Human behavior can be analyzed using Eye tracker. Thus, it is used for revealing the cognitive processes for object identification. Cognitive process is the mental ability for identification of what our eyes see. Vision with 20/20 sometimes may not reveal the purpose. In this study, ambiguous images are taken to observe the cognitive process in participants. During the perception of an object, a participant uses goal-directed search for identifying various objects. Dense gaze coordinates provide the region of interests and are considered as the target regions for object identification in ambiguous images. These data are used to develop cognitive models for identification of ambiguous images. Features such as, eye fixation, pupil diameter, fixation durations, moments of inertia, and polar moments are used for developing the cognitive model. Three different feature selection methods along with six different classifiers are used for the task of classification. The selection of a subset of features using hypothesis testing performed well, compared to principal component analysis based dimensionality reduction method. This study could be used in detecting whether a participants is lying or not while perceiving an ambiguous image. [ABSTRACT FROM AUTHOR]

To identify NLOS propagations for a high-resolution wireless localization system, a strong NLOS classifier is proposed via a machine learning algorithm named AdaBoost. Signal features are extracted to characterize the distinctions between LOS and NLOS, based on which a series of weak learners are primarily given. By training and re-sampling data iteratively, weak learners with the minimal classification error are selected and the weighted sum is boosted into a strong classifier. Simulations demonstrate that the proposed AdaBoost-based NLOS classifier can improve NLOS identification performance, leading to reduced positioning errors. [ABSTRACT FROM AUTHOR]

Owing to many appealing properties, neural networks provide a natural basis for solving different kinds of problems. The performance of neural networks greatly depends on whether they can provide appealing solutions to the problems of the parameter learning (i.e., the connecting weights in each layer) and the structure learning (i.e., the network structure). These two kinds of learning can be performed simultaneously or separately. In this paper, we proposed the Jacobian matrix-based learning machine (JMLM) to provide an appealing solution to the aforementioned two kinds of learning. The network structure of a JMLM can be incrementally constructed and a Jacobian-matrix-based learning method is proposed to efficiently estimate the corresponding network parameters. Furthermore, we can provide physically meaningful explanations to help human analyzers to make decisions based on the parameters embedded in a trained JMLM. One 2-D artificial data set, one benchmark medical data set, and an intensive care unit survival prediction data set were used for demonstrating the performance of the proposed JMLM.

The existing noise detection methods required the classifiers or distance measurements or data overall distribution, and 'curse of dimensionality' and other restrictions made them insufficiently effective in complex data, e.g., different attribute weights, high-dimensionality, containing feature noise, nonlinearity, etc. This is also the main reason that the existing noise filtering methods were not widely applied and formed an effective learning framework. To address this problem, we propose here a complete and efficient random forest method (CRF) specifically for the class noise detection by simulating the grid generation and expansion. The CRF is not based on distance measures or overall distribution or classifiers; besides, the voting mechanism makes it able to effectively process datasets containing feature noise. Furthermore, we introduce CRF based class noise filtering learning framework (CRF-NFL) and derive its mathematical model. The framework is then applied to many widely used classifiers including some state-of-the-art algorithms, e.g., k-means tree, GBDT, and XGBoost. Moreover, its parallelized is designed for large-scale data. The CRF-NFL show much better generalizability than the conventional classifiers and the relative density-based method, which is the most effective noise filtering method as far as we know. All research has formed an open source library, called CRF-NFL: http://www.cquptshuyinxia.com/CRF-NFL.html. [ABSTRACT FROM AUTHOR]