Search

In supervised learning, labeled data are provided as inputs and then learning is used to classify new observations. Error tolerance should be guaranteed for classifiers when they are employed in critical applications. A widely used type of classifiers is based on voting among instances (referred to as single voter classifiers) or multiple voters (referred to as ensemble classifiers). When the classifiers are implemented on a processor, Time-Based Modular Redundancy (TBMR) techniques are often used for protection due to the inflexibility of the hardware. In TBMR, any single error can be handled at the cost of additional computing either once for detection or twice for correction after detection; however, this technique increases the computation overhead by at least 100%. The Voting Margin (VM) scheme has recently been proposed to reduce the computation overhead of TBMR, but this scheme has only been utilized for k Nearest Neighbors (kNNs) classifiers with two classes. In this paper, the VM scheme is extended to multiple classes, as well as other voting classifiers by exploiting the intrinsic robustness of the algorithms. kNNs (that is a single voter classifier) and Random Forest (RF) (that is an ensemble classifier) are considered to evaluate the proposed scheme. Using multiple datasets, the results show that the proposed scheme significantly reduces the computation overhead by more than 70% for kNNs with good classification accuracy and by more than 90% for RF in all cases. However, when extended to multiple classes, the VM scheme for kNNs is not efficient for some datasets. In this paper, a new protection scheme referred to as k + 1 NNs is presented as an alternative option to provide efficient protection in those scenarios. In the new scheme, the computation overhead can be further reduced at the cost of allowing a very low percentage of errors that can modify the classification outcome. [ABSTRACT FROM AUTHOR]

In future 5G user-centric ultra-dense networks (UUDN), demands of high data rate and high spectrum efficiency are effectively met by dual connectivity (DC) technology. However, due to huge increase of base stations (BSs) and mobile users (MUs), it becomes difficult for BSs to quickly and precisely select the codeword and provide DC to MUs. Hence, different from some traditional methods, this correspondence paper aims to improve the network performance using the method of machine learning. First, we model the random distribution of BSs by homogeneous Poisson point processes, where each MU is served by millimeter-wave channel. Second, the probabilities that macro cell BS or small cell BS serves the MU are further derived to get the average sum rate (ASR) in UUDN. Third, inspired by ML, we utilize an iterative support vector machine (SVM) classifier to select the codewords of BSs, with sequential minimal optimization (SMO) algorithm used for training all link samples in UUDN. Then, an iterative SVM-SMO classification algorithm is proposed to achieve a highly efficient performance of DC, where the convergence and complexity are also discussed. The sample training and simulation at last are evaluated by Google TensorFlow. The simulation verified that our proposed algorithm gets a higher ASR than the traditional channel estimation based algorithm. In addition, the results also show a lower computational complexity can be achieved by the proposed algorithm as well. [ABSTRACT FROM AUTHOR]

In this paper, blind signal classification and detection in a non-orthogonal multiple access (NOMA) system are explored. Since a NOMA scheme superposes the multiple-user signals within nonorthogonal resources, classical modulation classification methods used in orthogonal multiple access (OMA) systems are not sufficient to process the superposed NOMA signal. NOMA receivers require information about the multiple access schemes such as modulation order and need interference cancellation of the co-scheduled user's signal; therefore, a NOMA system causes more high-layer signaling overheads than OMA during packet scheduling. Blind detection algorithms used for multiplexing information are considered to be possible solutions; however, they pose various challenges and could cause performance loss while performing blind modulation classification in the order of OMA/NOMA classification, co-scheduled user's modulation classification, and classification of the signal, due to the necessity for successive interference cancellation. To improve the performance of blind detection, we propose a NOMA transmission scheme that applies phase rotation to data or pilot symbols depending on the NOMA multiplexing format, as an aid to the blind detection. The proposed classification algorithm can implicitly provide essential information on NOMA multiplexing without the need for any extra high layer signaling or resources. The performance improvement is verified through simulation studies, and it is found that the proposed algorithm provides a gain of more than 1 dB compared to the existing blind signal classification methods and shows almost equivalent performance as the genie information scheme. [ABSTRACT FROM AUTHOR]

A rigorous model for automatic modulation classification (AMC) in cognitive radio systems is proposed in this paper. This is achieved by exploiting the Kalman filter (KF) integrated with an adaptive interacting multiple model (IMM) for resilient estimation of the channel state information. A novel approach is proposed, in adding up the square root singular values of the decomposed channel using the singular value decompositions algorithm. This new scheme, termed Frobenius eigenmode transmission, is chiefly intended to maintain the total power of all individual effective eigenmodes, as opposed to keeping only the dominant one. The analysis is applied over multiple-input multiple-output (MIMO) antennas in combination with a Rayleigh fading channel using a quasi-likelihood ratio test algorithm for AMC. The expectation-maximization is employed for recursive computation of the underlying estimation and classification algorithms. Novel simulations demonstrate the advantages of the combined IMM-KF structure when compared to the perfectly known channel and maximum likelihood estimate, in terms of achieving the targeted optimal performance with the desirable benefit of less computational complexity loads. [ABSTRACT FROM AUTHOR]

One of the serious security breaches that threatens today's smart technologies is the broadcasting of false alarms. These alarms may severely affect road management systems. Vehicular networks suffer from this type of threat because of the opacity and the on-purpose broadcasting of false contents which occur frequently. Modern cloud-fog communication architectures force the organization of hierarchical communications, resulting in more robust governance of disseminated contents. This paper proposes a prediction technique that is based on a binary classification using hidden Markov model (HMM). Our contribution in this research area are as follows: 1) modeling fog data features to enforce organization of communication, 2) proposing a technique for the automatic modeling of HMM parameters, which facilitate the precise prediction of false contents, and 3) proposing a dissemination algorithm for fast communication. We conducted several experiments to measure and compare the performance of the proposed model with the state-of-the-art classification algorithms. We also performed a misuse analysis to measure the ability of our proposed model to detect false alarms in synthesis and real-world datasets. In addition, we conducted some further analysis to demonstrate the sensitivity of our methodology to cutoff points. The results showed that our HMM-based technique provides promising prediction accuracy and significantly outperforms existing classification techniques in detecting false-alarm contents. [ABSTRACT FROM AUTHOR]

An in-depth study on driving habits and personalized driving-assistance systems is conducive to the realization of vehicle safety and intelligent driving. In this paper, we present a personalized vehicle lane-change assistance system integrated with a driver-behavior identification strategy. First, the driver-behavior data-acquisition system is designed and established. Based on this, the input data of different kinds of drivers along with vehicle signals are collected under typical working conditions. The drivers are classified utilizing factor analysis and a fuzzy c-means clustering algorithm, and the identification of driver behavior is realized using a backpropagation neural network optimized by a particle swarm optimization algorithm. Then, personalized warning, planning, and control systems are designed for lane changing. The proposed personalized lane-change assistance system can provide more personalized recommendations to the drivers, increasing the potential for more widespread acceptance and use of advanced driver-assistance system. Finally, the correctness of the proposed personalized lane-change system is evaluated by conducting computer simulations and a driver-in-the-loop- simulation under various conditions. And the results show that the lane-change assistance system based on driver behavior can meet the driving needs of different drivers without sacrificing safety. [ABSTRACT FROM AUTHOR]

The recognition of driver's braking intensity is of great importance for advanced control and energy management for electric vehicles. In this paper, the braking intensity is classified into three levels based on novel hybrid unsupervised and supervised learning methods. First, instead of selecting threshold for each braking intensity level manually, an unsupervised Gaussian mixture model is used to cluster the braking events automatically with brake pressure. Then, a supervised Random Forest model is trained to classify the correct braking intensity levels with the state signals of vehicle and powertrain. To obtain a more efficient classifier, critical features are analyzed and selected. Moreover, beyond the acquisition of discrete braking intensity level, a novel continuous observation method is proposed based on artificial neural networks to quantitative analyze and recognize the brake intensity using the prior determined features of vehicle states. Experimental data are collected in an electric vehicle under real-world driving scenarios. Finally, the classification and regression results of the proposed methods are evaluated and discussed. The results demonstrate the feasibility and accuracy of the proposed hybrid learning methods for braking intensity classification and quantitative recognition with various deceleration scenarios. [ABSTRACT FROM AUTHOR]

In urban environments, Global Positioning System (GPS) signals are often reflected or blocked by buildings causing multipath effects and Non-Line-Of-Sight (NLOS) reception. These effects degrade GPS positioning performance. Although improved receiver and antenna design can reduce the multipath effect to some extent, NLOS elimination is limited due to sensor cost. Measurements based modelling methods have shown promise to reduce NLOS, while positioning accuracy is limited by the correct classification of the signal reception types. These limitations are addressed in this paper by developing an integrated algorithm using matching, Particle Swam Optimized Adaptive Neuro Fuzzy Inference System (PSO-ANFIS) and a spatial city model. The user's location is determined by matching the high accuracy signal reception type determined by PSO-ANFIS with satellite visibility of each candidate position determined by the ray-tracing and a 3D city model. The results from a field test in an urban area in Taiwan show that the PSO-ANFIS based algorithm achieves a classification accuracy of 96%, 88%, and 91% for LOS, multipath and NLOS signals, respectively, which is superior to the other classification algorithms. The final horizontal positioning accuracies in terms of the Root Mean Square Error (RMSE) are 3.20 m using a commercial receiver and 1.88 m using a geodetic receiver. The proposed algorithm has consistent improvements of 88% and 70% respectively compared to the traditional Weighted Least Square (WLS) based positioning method, and is the best compared to the other state-of-the art methods with the improvements ranging from 15% to 77% depending on the test conditions. [ABSTRACT FROM AUTHOR]

Blind modulation classification (MC) is an integral part of designing an adaptive or intelligent transceiver for future wireless communications. However, till date, only a few works have been reported in the literature for blind MC of orthogonal frequency division multiplexing (OFDM) system over frequency-selective fading environment. In this paper, a blind MC algorithm has been proposed and implemented over National Instruments (NI) testbed setup for linearly modulated signals of OFDM system by using discrete Fourier transform (DFT) and normalized fourth-order cumulant. The proposed MC algorithm works in the presence of synchronization errors, i.e., frequency, timing, and phase offsets and without the prior information about the signal parameters and channel statistics. To nullify the effect of timing offset in the feature extraction process, a statistical average has been taken over OFDM symbols after introducing uniformly distributed random timing offsets in each of the OFDM symbols. In this work, we have classified a more extensive pool of modulation formats for OFDM signal, i.e., binary phase-shift keying (BPSK), quadrature PSK (QPSK), offset QPSK (OQPSK), minimum shift keying (MSK), and 16 quadrature amplitude modulation (16-QAM). Classification is performed in two stages. At the first stage, a normalized fourth-order cumulant is used on the DFT of the received OFDM signal to classify OQPSK, MSK, and 16-QAM modulation formats. At the second stage, first we compute the DFT of the square of the received OFDM signal and then a normalized fourth-order cumulant is used to classify BPSK and QPSK modulation formats. The success rate of the proposed MC algorithm is evaluated through analytical and Monte Carlo simulations and compared with existing methods. Finally, the work is validated by providing an experimental setup on NI hardware over an indoor propagation environment. [ABSTRACT FROM AUTHOR]

Turning behavior classification of oncoming vehicles is a critical step toward high-level scene understanding and motion planning tasks for autonomous vehicles. In this paper, we propose a new method for classifying an oncoming vehicle's turning behavior at intersection using 3D LIDAR, i.e., a Velodyne HDL-64E LIDAR, which collects rich point clouds from the surrounding environment. First, the vehicle pose is measured via a novel rectangle-based multi-frame fitting method applied to the vehicle point cloud extracted by a data pre-process procedure. Then, a Kalman filter integrates these measured vehicle poses to estimate and update vehicle kinematic parameters. Next, a hidden Markov model and Bayesian filtering are combined to recognize the vehicle's turning behavior using the heading angle and the acceleration components of the kinematic model. Real point clouds from intersection environments are used to train and validate the proposed turning behavior classification algorithm in the experiments. The results of the performance analysis and classification time measurements confirm that the proposed method can recognize the behavior of the oncoming vehicle at intersections correctly at an early stage of entering an intersection. [ABSTRACT FROM AUTHOR]

Support vector regression (SVR) algorithms have been applied to the identification of many nonlinear dynamic systems due to their excellent approximation and generalization capability. However, the standard SVR algorithm involves an iterative optimization process, which is often computationally expensive and inefficient. For applications such as the battery state-of-health (SOH) monitoring, where the identification algorithm needs to be applied repeatedly for multiple cells because of the variation in model dynamics (due to battery aging and cell-to-cell difference), the computational burden could pose difficulties for real-time or onboard implementation. In this paper, the battery V-Q curve identification problem for SOH monitoring is studied. Based on experimental battery aging data, we develop a model parametrization and adaptation framework utilizing the simple structure of SVR representation with determined support vectors (SVs) so that the model parameters can be estimated in real time. Through mathematical analysis and simulations using a mechanistic battery aging model, it is shown that the SVs of the battery models stay invariant, even when the batteries age or vary. The invariance of the SVs is verified using experimental aging data. Consequently, the resulting model for the battery V-Q curve can be directly incorporated into the battery management system (BMS) and adapted online for SOH monitoring. Moreover, the general characteristics of the data that could maintain the SVR invariance are identified. The proposed automated model parametrization process (via an optimization algorithm) can be extended to nonlinear dynamic systems with the given properties. [ABSTRACT FROM PUBLISHER]

In this paper, an engine fault detection and classification technique using vibration data in the crank angle domain is presented. These data are used in conjunction with artificial neural networks (ANNs), which are applied to detect faults in a four-stroke gasoline engine built for experimentation. A comparative study is provided between the popular backpropagation (BP) method, the Levenberg–Marquardt (LM) method, the quasi-Newton (QN) method, the extended Kalman filter (EKF), and the smooth variable structure filter (SVSF). The SVSF is a relatively new estimation strategy, based on the sliding mode concept. It has been formulated to efficiently train ANNs and is consequently referred to as the SVSF-ANN. The accuracy of the proposed method is compared with the standard accuracy of the Kalman-based filters and the popular BP algorithms in an effort to validate the SVSF-ANN performance and application to engine fault detection and classification. The customizable fault diagnostic system is able to detect known engine faults with various degrees of severity, such as defective lash adjuster, piston chirp (PC), and chain tensioner (CT) problems. The technique can be used at any dealership or assembly plant to considerably reduce warranty costs for the company and manufacturer. [ABSTRACT FROM PUBLISHER]

In this paper, we present an efficient allocation of power and channel so that power-drained primary users can harvest energy from secondary systems through wireless power transfer, and in return, the secondary system can have full access to the licensed spectrum for a certain length of time. Considering partial channel state information between primary and secondary systems, we formulate an optimization problem consisting of probabilistic constraint. By introducing a confidence level and using support vector machine, we convert the probabilistic constraint to deterministic form. We also implement a particle swarm optimization technique to solve the resource allocation optimization problem, while guaranteeing the required data rate for both primary and secondary systems. The results verify that the proposed scheme enables a higher number of primary users to efficiently harvest energy, and it also increases total primary capacity. [ABSTRACT FROM AUTHOR]

Cognitive small-cell networks have great potential in improving spectrum efficiency and mitigating inter-cell interference. Comprehensively classifying the channel states in cognitive small-cell networks is important for efficiently reusing the spectrum bands that are licensed to a macrocell. In this paper, we investigate channel state classification in cognitive small-cell networks with multilevels of transmission powers, including occupation detection of spectrum bands and transmission power classification of a macrocell base station (MBS). Specifically, two scenarios including a priori known signaling features and unknown signaling features are both studied. For the former scenario, we propose an optimal spectrum sensing and power classification algorithm, based on coherent classification, to achieve accurate sensing performance by fully exploiting the inherent information of the signaling features. Optimal sensing threshold and decision regions are derived for detecting and classifying the transmission power of the MBS. For the scenario without signaling features, a generic spectrum sensing and power classification algorithm is proposed based on noncoherent classification with low implementation complexity. A new performance metric, i.e., classification probability, is introduced to comprehensively evaluate the classification capability of the proposed algorithms. Finally, extensive simulations are provided to verify the proposed algorithms. [ABSTRACT FROM AUTHOR]

The potential of plug-in hybrid electric vehicles (PHEVs) to operate in electric and hybrid modes and their ability to supplement the energy storage off the grid have made them a front-runner in alternative fuel vehicle development. However, there is currently no widely accepted standard classification that provides an accurate comparison of PHEVs. This paper presents a novel classification for PHEVs: “Pihef: Plug-In Hybrid Electric Factor.” Pihef is the average ratio of the energy provided by the grid to the sum of the energy provided by the grid and fuel. The viability of Pihef is demonstrated via comprehensive simulations and a sensitivity analysis. In addition, the relationship between Pihef and emissions, efficiency, hybridization, and electric range is developed. [ABSTRACT FROM PUBLISHER]

Space–time block code (STBC) classification algorithms have recently received growing attention in academia and industry. In addition to their use in the context of military operations, these algorithms found civilian applications in reconfigurable systems, such as software-defined and cognitive radios. The previously reported single-carrier-based STBC classification algorithms are limited to frequency-flat fading channels; however, the wireless channels are typically frequency selective. This paper exploits the dispersive nature of the frequency-selective fading channels to classify Alamouti (AL) and spatial multiplexing (SM) STBCs over such channels. We show that the cross-correlation function of two different received signals for AL exhibits peaks at a particular set of time lags, whereas that for SM does not. Furthermore, we develop a maximum-likelihood classification algorithm. This requires channel knowledge, which may be unavailable in some scenarios such as radio environment awareness in cognitive radios. To avoid this requirement, we also propose a new classification algorithm based on the false alarm rate. Monte Carlo simulations are conducted to demonstrate the performance of the proposed algorithms. [ABSTRACT FROM PUBLISHER]

Plug-in hybrid electric vehicles (PHEVs) offer an immediate solution for emissions reduction and fuel displacement within the current infrastructure. Targeting PHEV powertrain optimization, a plethora of energy management strategies (EMSs) have been proposed. Although these algorithms present various levels of complexity and accuracy, they find a limitation in terms of availability of future trip information, which generally prevents exploitation of the full PHEV potential in real-life cycles. This paper presents a comprehensive analysis of EMS evolution toward blended mode (BM) and optimal control, providing a thorough survey of the latest progress in optimization-based algorithms. This is performed in the context of connected vehicles and highlights certain contributions that intelligent transportation systems (ITSs), traffic information, and cloud computing can provide to enhance PHEV energy management. The study is culminated with an analysis of future trends in terms of optimization algorithm development, optimization criteria, PHEV integration in the smart grid, and vehicles as part of the fleet. [ABSTRACT FROM AUTHOR]

The problem of identifying digital amplitude-phase modulations under additive noise is addressed within the theory of empirical Bayes deconvolution. The presented methods employ parametric models in the observation and signal constellation spaces. The model parameters are estimated using the received samples and then substituted into the respective models to obtain the estimate for the signal constellation, from which the decoding of the received samples can be accomplished. The proposed framework can be used to construct a modulation dictionary for an unknown transmitter prior to employing any hypothesis testing-based classification algorithm. [ABSTRACT FROM AUTHOR]