Search

Named data networking (NDN) aims to change the traditional content delivery method and caching by router nodes caching and participating in forwarding. NDN-caching can reduce the expected flood of global data traffic by providing cache storage at intermediate nodes for transmitted content objects, making data broadcasting in an efficient way. In this paper, a novel caching strategy based on entropy weighting method and TOPSIS is proposed for efficient content dissemination to improve NDN’s in-network caching performance. Firstly, the consumer’s request process is modeled by the entropy weighting method and TOPSIS to obtain the best cache node for the cache object according to the real-time status of the node. Secondly, two cache replacement algorithms (composed of an active cache replacement algorithm and a passive cache replacement algorithm) are proposed to reduce the cache redundancy on the delivery path and improve the utilization of data packets in the nodes. Finally, an effective cache mechanism and data packet migration scheme are proposed to further improve the cache performance according to the different types of cache nodes. The performance evaluation shows that the proposed scheme performs better in terms of cache hit rate, latency, and link load compared with some existing strategies.

Energy consumption has risen to be a bottleneck in wireless sensor networks. This is caused by the challenges faced by these networks due to their tiny sensor nodes that have limited memory storage, small battery capacity, limited processing capability, and bandwidth. Data compression has been used to reduce energy consumption and improve network lifetime, as it reduces data size before it can be forwarded from the sensing node to the sink node in the network. In this paper, a survey and comparison of currently available data compression techniques in wireless sensor networks are conducted. Suitable sets of criteria are defined to classify existing data compression algorithms. An adaptive lossless data compression algorithm (ALDC) is analyzed through MATLAB coding and simulation from the reviewed data compression techniques. The analysis aims to discover strategies that can be used to reduce the amount of data further before it is transmitted. From this analysis, it was discovered that encoding residue samples, rather than raw data samples, reduced the bitstream from 112 bits to a range of 30 to 36 bits depending on the sample block sizes. The average length of data samples to be passed to the encoder was minimized from the original 14 bits per symbol to 1.125 bits per symbol. This demonstrated a 0.875 code efficiency or redundancy. It resulted in an energy saving of 67.8% to 73.2%. This work further proposes a data compression algorithm that encodes the residue samples with fewer bits than the ALDC algorithm. The algorithm reduced the bitstream to 26 bits. The average length of the code is equal to the entropy of the data samples, demonstrating zero redundancy and an improved energy saving of 76.8% compared to ALDC. The proposed algorithm, therefore, shows improved energy efficiency through data compression.

Blood components such as hemoglobin, glucose, creatinine measuring are essential for monitoring one’s health condition. The current blood component measurement approaches still depend on invasive techniques that are painful, and uncomfortable for the patients. To facilitate measurement at home, we proposed a novel non-invasive technique to measure blood hemoglobin, glucose, and creatinine level based on PPG signal using Deep Neural Networks (DNN). Fingertip videos from 93 subjects have been collected using a smartphone. The PPG signal is generated from each video, and 46 characteristic features are then extracted from the PPG signal, its derivatives ( $1^{ \text {st}}$ and $2^{ \text {nd}}$ ) and from Fourier analysis. Additionally, age and gender are also included to feature because of the significant effects on hemoglobin, glucose, and creatinine. A correlation-based feature selection (CFS) using genetic algorithms (GA) has been used to select the optimal features to avoid redundancy and over-fitting. Finally, DNN based models have been developed to estimate the blood Hemoglobin (Hb), Glucose (Gl), and Creatinine (Cr) levels from the selected features. The approach provides the best-estimated accuracy of $R^{2} = 0.922$ for Hb, $R^{2} = 0.902$ for Gl, and $R^{2} = 0.969$ for Cr. Experimental aftermaths show that the proposed method is a suitable technique to be used clinically to measure human blood component levels without taking blood samples. This paper also reveals that smartphone-based PPG signal has a great potential to measure the different blood components.

In this article, we propose a depth data coding method by plane modeling. The plane modeling is a prediction method based on a plane estimation from depth pixels in a block. The plane modeling improves the intra-picture prediction for depth data comparing with intra modes of conventional coding standards. The plane modeling coefficients, which are the information about the estimated plane, are required to be provided during the depth data coding. The plane modeling coefficients are predicted from neighboring depth pixels. A prediction error of the plane modeling coefficients is calculated through the plane modeling for the selected pixels. If the prediction error is below a certain threshold, then the plane modeling coefficients are applied to the plane modeling for the block. From the simulation results, we verify that the proposed method achieves up to 6.76% bit rate saving in the same coding distortion condition compared to VVC test model.

A high-quality data transmission scheme in wireless body area networks is implemented by a joint source-channel coding (JSCC) system with $M$ -ary differential chaos shift keying modulation over an on-body channel. Current DP-LDPC code pairs, which have shown good bit-error ratio performance over additive white Gaussian noise channel and Rayleigh fading channel, cannot perform well over on-body channel. In this paper, the DP-LDPC codes are redesigned for the JSCC system with new methods depending on the on-body channel characteristics. A protomatrix dimension-restrictive searching algorithm for DP-LDPC codes in the JSCC system is proposed according to the on-body channel bandwith and source statistic, which is effective to predesign the dimension of the code pair for a new JSCC system. In this way, the source-channel code rates are appropriately allocated to leave more source redundancy for improving the error floor. Moreover, the joint protomatrix is reconstructed into two kinds of compact structures by simplifying the edge linking relationships and modifying joint protograph extrinsic information transfer algorithm, which decreases the induction of searching entries to further accelerate the code searching speed. The searched code pairs show better bit-error ratio performances compared with existing DP-LDPC codes. The power consumption in the new joint coding system is lower than the separate coding system in WBAN on the physical layer.

This paper is about the estimation of the cyber-resilience of CPS. We define two new resilience estimation metrics: $k$ -steerability and $\ell $ -monitorability. They aim at assisting designers to evaluate and increase the cyber-resilience of CPS when facing stealthy attacks. The $k$ -steerability metric reflects the ability of a controller to act on individual plant state variables when, at least, $k$ different groups of functionally diverse input signals may be processed. The $\ell $ -monitorability metric indicates the ability of a controller to monitor individual plant state variables with $\ell $ different groups of functionally diverse outputs. Paired together, the metrics lead to CPS reaching $(k,\ell)$ -resilience. When $k$ and $\ell $ are both greater than one, a CPS can absorb and adapt to control-theoretic attacks manipulating input and output signals. We also relate the parameters $k$ and $\ell $ to the recoverability of a system. We define recoverability strategies to mitigate the impact of perpetrated attacks. We show that the values of $k$ and $\ell $ can be augmented by combining redundancy and diversity in hardware and software, in order to apply the moving target paradigm. We validate the approach via simulation and numeric results.

For multidimensional data, Space-Filling Curves (SFCs) have been used to improve the execution time of spatial data queries. However, their effect on compression, when used to reorder the uncompressed values, is known to a lesser extent. We investigate the impact of three SFCs on Shuttle Radar Topographic Mission (SRTM) elevation data and Square-Kilometre Array telescope (SKA) radio-astronomy data: two types of datasets to which SFCs have not been extensively applied, within a compression context. This work contributes to the understanding of how such reorderings impact compression performance and affect different compression schemes and preprocessing techniques through their use. We show empirical results from combining eight common compression schemes, the Z-Order, Gray-Code, and Hilbert space-filling curves, and the bitwise preprocessing technique BitShuffle. The Hilbert Curve consistently outperforms the other orderings for the SRTM dataset though the mapping implementation incurs a significant speed penalty. However, the Z-Order and Gray-Code Curves are best for the SKA dataset. Through an analysis of the dataset autocorrelations, file-entropies, and block-entropies; we show that the SKA dataset’s dimensional bias is not exploited as much by the Hilbert Curve compared to the Z-Order and Gray-Code Curves. However, the Hilbert Curve is the most appropriate for the SRTM dataset as it can be modelled as isotropic and has a significantly higher level of local autocorrelation. BitShuffle is necessary to practically compress the SKA data, but does contribute to the compression performance of the SRTM dataset. These curves and BitShuffle are advantageous in reducing block-entropy values for such datasets.

MeanShift is one of the popular clustering algorithms and can be used to partition a digital image into semantically meaningful regions in an unsupervised manner. However, due to its prohibitively high computational complexity, a grid-based approach, called MeanShift++, has recently been proposed and succeeded to surprisingly reduce the computational complexity of MeanShift. Nevertheless, we found that MeanShift++ still has computational redundancy and there is room for improvement in terms of accuracy and runtime; thus, we propose an improvement to MeanShift++, named $\alpha $ -MeanShift++. We first attempt to minimize the computational redundancy by using an additional hash table. Then, we introduce a speedup factor ( $\alpha $ ) to reduce the number of iterations required until convergence, and we use more neighboring grid cells for the same bandwidth to improve accuracy. Through intensive experiments on image segmentation benchmark datasets, we demonstrate that $\alpha $ -MeanShift++ can run 4.1- $4.6\times $ faster on average (but up to $7\times $ ) than MeanShift++ and achieve better image segmentation quality.

Considering that large quantities of soil hyperspectral data include the redundancy and overlap of spectral information, the technology of selecting feature wavelength can effectively solve these problems, and improve the accuracy and stability of the soil organic matter (SOM) content retrieval model. Traditional methods of wavelength selection mainly attempt to establish the empirical relationship between reflectance and SOM contents, and the performance is directly related to the quality and representativeness of the “training data”. This study first distinguished the sensitive wavelength interval of SOM through the sensitivity analysis (SA) of the SOM to soil reflectance in radiative transfer model. Then sensitive wavelength points of SOM were ascertained using the successive projection algorithm (SPA): 468, 476, 496, 599, 775 and 900nm. Results show that SOM content can be estimated with high accuracy (root-mean-square error of prediction (RMSEP) 2) > 82.9%) by adopting the selected six wavelengths. Especially at 599nm, the accuracy of SOM content estimation is the highest (RMSEP: 0.176%, R2: 90.4%). Compared with traditional empirical wavelength selection methods, the wavelength selection based on the SA-SPA with the SOM radiative transfer model improves the generalizability and accuracy of the result. The research provides theoretical basis and technical support for the remote sensing retrieval of SOM, the development of rapid spectral instruments, and the bands setting of sensor instrument.

This article focuses on the estimation of the direction of Arrival based on the interpolation approaches of the sparse array. Such as the minimum redundancy array, co-prime array, and the nested array have attracted much attention, especially for a non-linear array antenna. We analyze the effect of antennas locations errors on overall estimation performance for deterministic and stochastic error models and to improve the estimator performance in the case of uncorrelated sources of signals in spatially-smoothed MUSIC scenarios. In two scenarios: we determine the Cramér-Rao bound in the case of the antennas locations are perfect, and in other cases, we take into account antennas locations errors. For the first scenario, the Cramér-Rao bound converges as the SNR increase. For the second scenario, the Signal-to-noise ratio increases to infinity, to achieve the same Cramér-Rao bound as in the first scenarios. There is a significant gap between the Cramér-Rao bound values in a situation where antennas locations errors are considered, and when other locations errors are not considered. This disparity shows that unknown antenna locations errors have a significant impact on the estimated achievable performance of the direction of Arriva of scattered linear matrices.

This paper investigates Kalman Filter-based Heuristic Ensemble (KFHE), which is a new perspective on multi-class ensemble classification with performance significantly better or at least as good as the state-of-the-art algorithms. We prove that the sample weight tuning method used in KFHE is a version of adaptive boosting, and the weight distribution does not change anymore and leads to redundant classifiers when the algorithm iterates enough times. This motivates us to select a sub-ensemble to alleviate the redundancy and improve the performance of the ensemble. An Ordering-based Kalman Filter Selective Ensemble (OKFSE) is proposed in this paper to select a sub-ensemble using the margin distance minimization approach. We demonstrate the effectiveness and robustness of OKFSE through extensive experiments on 20 real-world UCI datasets, and the statistical test shows that OKFSE significantly outperforms the state-of-the-art KFHE and clustering-based pruning methods on these datasets with 5% and 10% class label noise.

Recent years, multi-label feature selection has gradually attracted significant attentions from machine learning, statistical computing and related communities and has been widely applied to diverse problems from music recognition to text mining, image annotation, etc. However, traditional multi-label feature selection methods employ cumulative summation strategy to design methods, which suffers from the problem of overestimation the redundancy of some candidate features. Additionally, the cumulative summation strategy will lead to the high value of the goal function when one candidate feature is completed related with one or few already-selected features, but it is almost independent from the majority of already-selected features. To address these issues, we propose a new multi-label feature selection method named Feature Redundancy Maximization (FRM), which combines the cumulative summation of conditional mutual information with the `maximum of the minimum' criterion. Additionally, FRM can be rewritten as another form of multi-label feature selection method that employs interaction information as a measure of feature redundancy that obtains an accurate score of feature redundancy as the number of already-selected features increases. Finally, extensive experiments are implemented on fourteen benchmark multi-label data sets in comparison to six state-of-the-art methods. The experimental results demonstrate the superiority of the proposed method.

Since great redundancy of telemetry data of spacecraft, telemetry data compression is a good solution for the limited bandwidth and contact wireless links. It is important to obtain accurate data characteristic firstly. State-of-the-art machine learning methods work well on data mining and pattern recognition under conditions of the given test data set, which could be used as the available tools for post-event data processing and analysis, such as trend forecasting and outlier detection, but they have not provided the proper solution from the source on-board. In this paper, four base classes of the telemetry data are suggested and studied through the time series feature and information entropy analysis, then a new on-board lightweight self-learning algorithm named Classification Probability calculation - Window Step optimization (CP-WS) is proposed to obtain the class features and make the decision of each single parameter from the continuous discrete telemetry time series. Simulation results show that, our algorithm correctly classifies the simulation and real mission data into the appropriate base class with advantages of high classification accuracy as 100% and adaptive computational complexity from $O(L^{2})$ to $O(L)$ , which could be used in satellite on-board data compression for space-to-ground transmission, especially for the deep space explorers to save important status with less on-board storage space.

In this paper, we designed a novel framework for massive image panorama stitching, which aims to resolve image redundancy, alignment error accumulation and perspective distortion accumulation of the stitching process. First, an iterative method is designed to filter the redundancy images by analysing the similarity relation of adjacent images. Second, to reduce alignment error accumulation, the weight topology graph of filtered images is constructed, which is employed to find the optimal reference image which closes to the central geometrically by the Floyd-Warshall algorithm. Finally, a two-step global alignment strategy is designed to initial align images and perform shape optimization, the first step is that filtered images employ the similarity model to roughly align group by group, the second step is to perform shape optimization further through refining all alignment parameters by the homography model under the anti-perspective energy term, which aims to obtain an optimal solution by balancing the alignment accuracy and the global consistency. Compared with the state-of-the-art methods, the proposed method successfully reduces image redundancy while improving the alignment and reducing perspective distortion for massive image panorama stitching.

Reversible data hiding in encrypted image (RDHEI) is a technique that can provide the security and invisibility of their own information during the acquisition and sharing of them among multi-users. Among which, parametric binary tree labelling (PBTL) is a novel technique designed to serve for high-capacity RDHEI. However, considering the local smoothness of the image, its potential redundancy room has not been fully explored. In this paper, we propose an improved PBTL-RDHEI scheme (IPBTL-RDHEI in short). In IPBTL-RDHEI, an adjusting pixel modulation strategy is considered to reduce the probability of the overflow of pixels, making more embeddable pixels usage to carry secret data. Moreover, a block-wise rearrangement and compression (BRC) mechanism is introduced to decrease the length of the required auxiliary information. Experimental results confirm that the proposed scheme is able to achieve an average embedding rate as large as 1.827 bpp when the block size is set to $3\times 3$ .

Human action recognition plays a key role in human-computer interaction in complex environments. However, similar actions will lead to poor feature sequence extraction and result in a reduction in recognition accuracy. This paper proposes a method (Action-Fusion: Multi-label subspace Learning (MLSL)) from depth maps called Depth Sequential Information Entropy Maps (DSIEM) and skeleton data for human action recognition in multiple modal features. The DSIEM describe the spatial information of human motion with information entropy, and describe the temporal information through stitching. DSIEM can reduce the redundancy of depth sequences and effectively capture spatial motion states. MLSL studies the relationship between different modalities and the inherent connection between different labels. The method is evaluated on three public datasets: Microsoft action 3D dataset (MSR Action3D), University of Texas at Dallas-multimodal human action dataset (UTD-MHAD), UTD MHAD- Kinect Version-2 (UTD-MHAD-Kinect V2). Experimental results show that the proposed MLSL model obtains new state-of-the-art results, including achieving the average rate of the MSR Action3D to 93.55%, the average rate of the UTD-MHAD to 88.37% and the average rate of the UTD-MHAD-Kinect V2 to 90.66%.

The C4.5 algorithm has three shortcomings: the wide range of candidate segmentation threshold sequences for continuous attributes, the comprehensive influence of different attributes and local subsets under the same attribute, and the inter-attribute redundancy. When dealing with continuous attributes, sampling and threshold supplement processing near the transition boundary of the attribute interval corresponding to the adjacent different categories are performed for narrowing the range of candate segmentation threshold sequences. By adding standardizing Euclidean distance of the attribute global and local factors to represent attribute weight, the calculation of C4.5 information gain is otpimized. And averaging Gini index of other attributes and adding correction factor, the influence of redundancy between attributes is greatly decreased. The overall average improvement range of the base classifier and the bagging integration classifier is 0.6%~2.1% and 0.7% ~ 2.7%, respectively, which shows that this integration model can improve the classification accuracy and also validate its feasibility and reliability.

L-shaped array (LA) is traditionally used as two one-dimensional uniform linear arrays (ULAs) for two-dimensional direction of arrival (DOA) estimations. With elevation and azimuth angles estimated separately, the detectable source number is strictly limited by the number of elements in the ULAs. To increase the detection ability, in this paper, we propose to exploit the diff-sum co-array (DSCA) concept on the L-shaped array. As such, a virtual element in the quadrants of DSCA can be generated by only one pair of sensors in the LA. Therefore, it can acquire much higher number of degrees of freedom for DOA estimation. To further reduce the redundancy of elements on the horizontal and vertical axes of DSCA, an element moving strategy is proposed. It is proved that the DSCA can remain intact with any one or more physical sensors relocated to the centrosymmetric positions. Based on this strategy, a generalized L-shape array concept and a new configuration referred to as generalized L-shaped array with odd-even locations (GLA-OEL) are developed. For GLA-OEL, the number of sensor pairs with small spacing is as small as one, which dramatically reduces the mutual coupling of adjacent elements. Besides, with the same number of sensors, the GLA-OEL can generate much larger virtual uniform rectangular array in the DSCA. Simulations verify the superiorities of the proposed array in terms of detection performance and estimation accuracy.

Recently, multilabel classification is of increasing interest in machine learning and artificial intelligence. However, the distances of samples in most Relief methods easily result in heterogeneous or similar samples abnormal when the distances are very large. Besides, the classification margin as a neighborhood radius for some reduction algorithms may be meaningless when the margin is too large. To overcome these drawbacks, this paper presents a multilabel feature selection method using the improved Relief and minimum redundancy maximum relevance (MRMR) based on neighborhood rough sets. First, the number of heterogeneous and similar samples is introduced to improve the label weighting method which can eliminate the influence of the large distances of samples. By combining with the new label weighting, the distances between the sample and its nearest-neighbor heterogeneous sample and between the sample and its nearest-neighbor similar sample are presented to develop a new feature weighting method. Second, the number of heterogeneous and similar samples continues to be used to improve the classification margin, thereby constraining the neighborhood radius, based on which the neighborhood approximation accuracy is constructed to effectively measure the uncertainty of samples in the boundary region and the completeness of knowledge. Third, by integrating with the new neighborhood approximation accuracy, two types of mutual information between features and labels and among features are proposed, and then the mutual information-based MRMR model is investigated to evaluate the significance of features. Finally, a multilabel feature selection algorithm is designed for improving the classification performance of multilabel data. Experimental results on thirteen public datasets illustrate the effectiveness of our developed algorithm that can select the significant features and achieve great performance for multilabel datasets.

Filters are the fastest among the different types of feature selection methods. They employ metrics from information theory, such as mutual information (MI), Joint-MI (JMI), and minimal redundancy and maximal relevance (mRMR). The determination of the optimal feature selection set is an NP-hard problem. This work proposes the engineering of the Genetic Algorithm (GA) in which the fitness of solutions consists of two terms. The first is a feature selection metric such as MI, JMI, and mRMR, and the second term is the overlapping-coefficient that accounts for the diversity in the GA population. Experimental results show that the proposed algorithm can return multiple good quality solutions that also have minimal overlap with each other. Numerous solutions provide significant benefits when the test data contains none or missing values. Experiments were conducted using two publicly available time-series datasets. The feature sets are also applied to perform forecasting using a simple Long Short-Term Memory (LSTM) model, and the solution quality of the forecasting using different feature sets is analyzed. The proposed algorithm was compared with a popular optimization tool `Basic Open-source Nonlinear Mixed INteger programming' (BONMIN), and a recent feature selection algorithm `Conditional Mutual Information Considering Feature Interaction' (CMFSI). The experiments show that the multiple solutions found by the proposed method have good quality and minimal overlap.

A multi-cyclic redundancy check (Multi-CRC) polar code construction algorithm is proposed in this paper to solve the error propagation problem of successive cancellation decoding for polar codes. In this algorithm, the information sequence is optimized into several segments to allow decoding errors to be corrected in time, minimizing the impact of error propagation. An improved multi-successive cancellation bit flipping (M-SCFlip) decoding algorithm is proposed to execute the bit flipping operation after CRC check-in each segment. In the low-SNR region, the proposed new multi-CRC polar code with successive cancellation list (SCL) decoding has a slight frame-error rate (FER) degradation compared with the original CRC polar code. With the M-SCFlip decoding algorithm developed in this paper, it achieves a better FER performance compared with the CRC polar code with successive cancellation (SC) and SCL ( $L\!=\!2$ ) decoding algorithms. In addition, it has a lower decoding delay and requires a lower memory space. For example, at a FER of 10−4 with the same code length and effective code rate, the proposed multi-CRC polar code with M-CFlip decoding achieves a 1.19 dB and 0.79 dB gains over existing CRC polar codes with the SC and SCL ( $L\!=\!2$ ) decoding algorithms, respectively.

As is well known, the prototype coprime arrays consist of two collinear subarrays, whose physical sensor number is coprime to each other. Moreover, their larger inter-element separations and physical array apertures make them have less mutual coupling than uniform linear arrays. Even so, there is still mutual coupling existing in them, which has an adverse effect on the direction of arrival estimation. To reduce the effects of mutual coupling, several improved coprime configurations based on the concept of difference and sum coarray are proposed in this paper, which can be obtained through three operations. The first one is to shift the prototype coprime array to obtain the translational coprime arrays, which can help to reduce the redundancy between the difference coarray and sum coarray. Second, flip one of the subarrays in the translational coprime arrays with zero points as the symmetric center to get the flipped coprime arrays, which can expand the physical array aperture and inter-element separations further. Finally, move one of the symmetrically placed sensors in the flipped coprime arrays to zero points. Then, the proposed arrays are derived. They have almost the same consecutive degrees of freedom (DOFs) as the prototype coprime array but dramatically reduced mutual coupling due to their larger physical array apertures and inter-element separations, where the virtual arrays are constructed by their respective difference and sum coarray. In the end, the numerical simulation results confirm the superiority of proposed arrays.

Due to the centralization of communication in the management of data generated by diverse Internet of Thing (IoT) devices, there is a lack of reliability when data is being transferred and stored. Among errors caused by various behaviors, Silent Data Corruption (SDC) error, owing to stealthy destruction without error prompt, is one of the most difficult data consistency problems in the storage system, whether it is a traditional multi-control, distributed storage, or public cloud storage. Nowadays, for SDC error detection, extracting instruction features to analyze vulnerabilities of programs or instructions has still not been fully explored. Literature generally just count the number of possible features, without mining the inter-characteristic of the instruction and correlation between them. Thus, we propose a method of SDC-causing Error Detection based on Support Vector Regression (SED-SVR) for fully exploiting the correlation between data features. Specifically, firstly, we extract instruction features based on the SDC vulnerability of program instructions by analyzing results of fault injections. Secondly, we establish the instruction SDC vulnerability prediction model based on SVR and propose our SED-SVR model. Thirdly, according to the predicted values of SDC vulnerability, we develop some solutions for faults tolerance of target programs by different granularity of instruction redundancy. The experimental results show that our SED-SVR has higher fault detection rate with lower performance overhead.

Nonnegative matrix factorization, as a classical part-based representation method, has been widely used in pattern recognition, data mining and other fields. However, the traditional nonnegative matrix factorization directly factoring decomposes the original data, and the original data often contains a lot of redundancy and noise, which seriously affect the subsequent processing of the data. In this work, we propose an adaptive graph regularization discriminant nonnegative matrix factorization (AGDNMF) for image clustering. The AGDNMF algorithm makes full use of local structure information and a small amount of label information. In AGDNMF, the local structure information can be more accurate and the label information can prevent the points with the same label from being merged into one point. These two items are combined into the objective function of NMF. In addition, we provide the update rules for the corresponding optimization functions and prove its convergence. A large number of experiments on different data sets show that the proposed algorithm has good clustering performance.

In modern chemical process control, the application of data-driven soft sensor has become increasingly extensive. Feature extraction is an important step in soft sensor. A novel feature extraction and integration method based on stacked autoencoders (SAE) and mutual information (MI)-weighted principle component analysis (PCA) was proposed to solve the loss of information on shallow depth features and original variables in neural network models. First, an SAE model was trained to extract the features of the original variables with varying depths. Second, through an MI indicator, the original variables and features with strong dependency on the outputs were selected. Then, MI was used to assign varied weights to the features and original variables, and the PCA method was used to remove any possible redundancy between the original variables and features of varying depths to obtain the principle components. Finally, the principle components were used to construct a regressor, such as a neural network. The model was first tested using the Boston housing dataset as a benchmark and then applied to the soft sensor of a constant top oil dry point. The proposed model achieved optimal results in terms of the root mean squared error and $r$ indicators in the experiments and was thus proved feasible and useful.

Unsupervised band selection plays an increasingly important role in a hyperspectral image (HSI) classification because of inadequate labeling samples. However, how to select more representative, less redundant, and informative band is an open problem. Recently, a fast density-peak-based clustering (FDPC) algorithm has been proposed. The FDPC chooses the cluster center through the local density and the intracluster distance of each point and ranks each point through a certain rule. For HSI band selection, the FDPC has the following problems. First, when calculating the local density of bands, the difference between bands is not considered, so the local density cannot better characterize the band distribution. Second, the ranking rule in FDPC is more inclined to select bands with larger density and intracluster distance values. However, some boundary bands with abundant information and low redundancy cannot be found. This paper proposes a local potential-based clustering algorithm for unsupervised hyperspectral band selection (LPC). The LPC algorithm improves the FDPC algorithm in three aspects of band selection. First, the local potential of each band is calculated according to the similarity of between bands, and the larger similarity has a greater effect on the local potential. The local potential can better characterize the band distribution. Second, the LPC proposes a weighted ranking rule, which integrates the three factors of local potential, intracluster distance and standard deviation to guide the algorithm to select more discriminative bands. In particular, some boundary bands with abundant information and low redundancy can be found. Finally, an effective method is designed to automatically select the appropriate number of bands. The experimental results on three real hyperspectral data sets demonstrate that the proposed method outperforms the other state-of-the-art methods.

Dimensionality reduction is used in microarray data analysis to enhance prediction quality, reduce computing time, and construct more robust models. In addition, the algorithm learning performance involves an expressive number of attributes (genes) relative to the classes (samples). Therefore, in this paper, we conducted a detailed comparison of two reduction methods, attribute selection and principal component analysis, to analyze gene expression data sets. Both reduction methods were employed in the pre-processing stage and then evaluated experimentally. Furthermore, we introduced a combination of consistency-based subset evaluation (CSE) and minimum redundancy maximum relevance (mRMR), which we referred to as CSE-mRMR, to improve classification efficiency. The results indicated a significant increase in classifier hit rates with both methods, compared to using all attributes. By employing cross-validation, attribute selection outperformed PCA consistently across classifiers and datasets, and CSE-mRMR demonstrated good classification performance in the data sets. Taken together, the literature and current results suggest that the attribute selection may be relevant in the analysis and future prediction of gene expression data sets.

Point cloud data from LiDAR sensors is the currently the basis of most L4 autonomous driving systems. Sharing and storing point clouds will also be important for future applications, such as accident investigation or V2V/V2X networks. Due to the huge volume of data involved, storing point clouds collected over long periods of time and transmitting point clouds in real-time are difficult tasks, making compression an indispensable step before storing or transmitting. Previous streaming point cloud compression methods, such as octree compression or video compression-based approaches, have difficulty compressing this data in real-time into very small volumes with low information loss. To reduce temporal redundancy efficiently and rapidly, in this paper we propose a real-time streaming point cloud data compression method using U-net. By utilizing raw packet data from LiDAR sensors, we can store 3D point cloud information losslessly in a 2D matrix, and convert streaming point cloud data into a video-like format. By designating some frames as reference frames and then using U-net to interpolate the remaining LiDAR frames, we can greatly reduce temporal redundancy. Our use of U-net was inspired by a video interpolation approach employed in another study. Noise in LiDAR data is a big issue which significantly affects network training and compression results. In this paper, we propose a padding strategy to alleviate the negative impact of this noise. As a result of these improvements, our proposed method can outperform octree compression, MPEG-based compression and our previously proposed SLAM-based compression method.

DenseNet has strong expressiveness in many computer vision tasks, but the complexity of its model makes it difficult to deploy to devices with limited computing resources. In this paper, we propose S-DenseNet, a compact model of DenseNet which makes the extracting features of DenseNet more comprehensively and reduces the parameter redundancy at the same time. Firstly, we investigate the parameter redundancy of DenseNet and show the possibility of compressing DenseNet. For example, we reduced its parameters by 10% as keeping in the same accuracy. Secondly, we propose an algorithm based on the Skyline method for converting standard convolution into group convolution automatically. It can choose the filter weights automatically and restore the standard convolution filter with a high sparsity one called S-Conv. Finally, we combine S-Conv with dense connections and formed a more effective network architecture, that is S-DenseNet. Experiments show that in Cifar10, Cifar100 and SVHN datasets, we use only 40% parameters and 20% FLOPs of DenseNet, and the accuracy is improved by 1%. Compared with the compression models of DenseNet and other CNNs, we use less complexity in the model and achieve higher or similar accuracy. On ImageNet dataset, compared with lightweight CNN models, traditional CNNs and their compression models, we achieve the higher or similar Top-1 accuracy with less complexity.

As a graphical model, Conditional Preference Networks (CP-nets) are used to describe the qualitative conditional preferences of attributes, and the structure learning plays an important role in the research of CP-nets. Different from traditional CP-nets structure learning methods, a Maximum Relevance Minimum Common Redundancy (mRMCR) algorithm based on the information theory and feature selection is proposed and discussed detailedly. Firstly, a mutual information solution formula on the preference database is established, it regards mutual information as mutual relation between one attribute and its feasible father set, which also avoids the calculation of conditional mutual information. Secondly, in order to make our graphical model include relevant, exclude irrelevant and control the use of redundant features, a formula for calculating mRMCR is designed. The mRMCR algorithm can measure the dependent relationship effectively and determine the causal relationship between variables, and can get the structure of CP-nets. Finally, the effectiveness of the algorithm is verified on the movie recommendation datasets. The experimental results show that the proposed mRMCR algorithm can not only obtain the causal relationship between variables quickly and effectively but also extract the feasible father set of each attribute and then obtain the topological structure of CP-nets.

Meliorating a priori stochastic model of Kalman filer (KF) is always challenging. To address this challenge, this paper simultaneously estimates and corrects the variance components for all of the process noise and measurement matrix ( $\boldsymbol {Q}$ & $\boldsymbol {R}$ ) by a posteriori variance-covariance components estimation (VCE) algorithm, which makes the most of the process noise residuals and measurement residuals and measurement redundancy contribution. Unsurprisingly, in the conventional error states-based integration mechanization, the stochastic model tuning is not easy for IMU because of the error measurements between the observables from inertial sensors and other aiding sensors. This research utilizes an unconventional multi-sensor integration strategy, in which a 3D kinematic trajectory model is deployed as the main part of system equation and the systematic errors of each IMU and the measurements of all sensors are individually modelled. Furthermore, the weights of measurements from each inertial sensor are defined on the basis of the posterior variances, so that we could properly distribute the function of each measurement in the fusion algorithm. A real dataset involving GPS and multiple IMUs is processed to validate the proposed posteriori VCE algorithm by applying the unconventional integration strategy.

High efficiency video coding (HEVC) is standard in the video compression field and performs well not only in video compression coding, but also in image compression. Regions of interest (ROIs) and just noticeable difference (JND), two human visual models, can accurately quantify human visual system (HVS) characteristics as pixel values. Using ROIs and JND to assist in evaluating image distortion can effectively reduce human visual redundancy and reflect authentic perceptual distortions. However, they are not readily applicable to the HEVC test model (HM) in the pixel domain. It is difficult to secure a suitable Lagrange multiplier $\lambda $ and quantization parameter (QP) for the JND model in particular. This paper proposes different solutions for the use of rate control (RC) or not where an appropriate $\lambda $ value is available for perceptual models. In RC, the proposed approach centers on a robust relationship between QP and achieved $\lambda $ . And we also established a bit allocation technique using the related $\lambda $ and expression for the RC model. Experimental results validate the rationality and effectiveness of the proposed method.

This paper proposes an optimized polar coded selective relay cooperation with an iterative threshold decision of pseudo posterior probability to solve the excessive redundancy of received signals and the high decoding complexity in a selective decode-and-forward (SDF) relay scheme. By sharing antennas of a relay node, a polar coded cooperation scheme is illustrated in a three-node relay transmission model, accompanied by an SDF protocol. Simultaneously, the SDF protocol is completed according to the channel status information at the relay node. In the logarithmic belief propagation algorithm, an expression of pseudo posterior probability-based stopping iteration threshold with independent and variable signal-to-noise ratios is derived to decide convergence condition and thus timely stop the iterations for low decoding complexity. And an extrinsic information transfer chart is also used to verify the completion of the iterative process. In addition, an information detection factor is introduced to check decoded bits correctly by the change of it, thereby reducing the average number of decoding iterations. It also accelerates the convergence speed of the BP algorithm by setting threshold and prejudgment, thereby improving the decoding efficiency. Simulation results show that the proposed polar coded relay cooperation scheme obtains over 2 dB performance gain, when the bit error rate (BER) is lower than 10-3, compared with the existing low-density parity-check codes related schemes. It also reduces the complexity of the BP algorithm by stopping the iteration of high-reliability nodes at the cost of performance loss within 0.5 dB. Therefore, the proposed scheme obtains both good BER performance and low complexity, which endows its good applications in cooperative wireless communications.

In this paper, we propose to use deep neural networks for image compression in the wavelet transform domain. When the input image is transformed from the spatial pixel domain to the wavelet transform domain, one low-frequency sub-band (LF sub-band) and three high-frequency sub-bands (HF sub-bands) are generated. Low-frequency sub-band is firstly used to predict each high-frequency sub-band to eliminate redundancy between the sub-bands, after which the sub-bands are fed into different auto-encoders to do the encoding. In order to further improve the compression efficiency, we use a conditional probability model to estimate the context-dependent prior probability of the encoded codes, which can be used for entropy coding. The entire training process is unsupervised, and the auto-encoders and the conditional probability model are trained jointly. The experimental results show that the proposed approach outperforms JPEG, JPEG2000, BPG, and some mainstream neural network-based image compression. Furthermore, it produces better visual quality with clearer details and textures because more high-frequency coefficients can be reserved, thanks to the high-frequency prediction.

Data hiding is a useful technology to protect secret data through the Internet. Exploiting modification direction (EMD)-based data hiding uses a group of $n$ cover-pixels to embed secret data. It achieves good image quality and high security level of resisting RS detection. However, the embedding capacity of EMD decreases fast when $n$ increases. In 2013, a Generalized EMD (GEMD) was proposed to improve the embedding capacity of EMD, which is always more than 1 bpp (bits per pixel), and keeps the good stego-image quality. In this paper, we propose an enhanced GEMD by dividing a group of $n$ cover-pixels into multiple groups. By this method, the embedding capacity can be further improved from GEMD. Our scheme maintains the good image quality and can resist the RS detection as well.

In this paper, a transmission redundancy optimization strategy for fountain codes that uses an adaptive filtering technique as well as recent results concerning the upper bound of the source packet erasure probability for fountain codes over finite fields is proposed. The performance of the proposed strategy was evaluated in static and simulated realistic underwater acoustic channel environments simulated by a recently proposed statistical underwater acoustic channel model. The goal of this investigation is to improve the communication efficiency of underwater acoustic sensor networks that use communication protocols based on fountain codes. The performance of the proposed strategy is compared with the recently proposed approaches of discrete stochastic approximation (DSA) and the joint power and rate control (JPR) schemes. The simulation results demonstrate that the strategy proposed in this paper performs better than the existing schemes in both static and simulated realistic environments. The proposed strategy is 4.75 times better than the DSA in the static convergence performance evaluation. Further analysis shows that in the presence of a simulated realistic environment, compared to DSA and JPR, the strategy proposed in this paper reduces the overall transmission redundancy by approximately 11.4% and 52.6%, respectively. It also improves the average communication efficiency by 1.3% and 12.5%, respectively. The average optimization error relative to optimal redundancy is improved by approximately 62.1% and 92.5%, respectively. These numerical results are obtained by employing the specific simulation parameters and the simulated environment mentioned in this paper.

In this paper, we propose a unified array geometry, dubbed generalized nested subarray (GNSA), for the underdetermined direction-of-arrival estimation. The GNSA is composed of multiple, identical subarrays, which can be a minimum redundancy array (MRA), a (super) nested array, a uniform linear array (ULA), or any other linear arrays with hole-free difference coarrays (DCAs). By properly design the spacings between subarrays, the resulting DCA of the GNSA can also be a hole-free (filled) ULA. When the subarray is an MRA and meanwhile its sensors’ positions also follow an MRA configuration, a nested MRA (NMRA) is constructed. This NMRA can provide up to $\mathcal {O}(M^{2}N^{2})$ degrees of freedom (DOFs) using only $MN$ physical sensors. In order to fully utilize the increased DOF, we develop a new DOA estimation algorithm, which consists of a dimensional reduction matrix to exploit the data of all virtual elements, a Toeplitz matrix to decorrelate the equivalent coherent sources, and a root-MUSIC method to mitigate the computational workload. This new algorithm can achieve better DOA estimation performance than traditional spatial smoothing MUSIC algorithm with lower computational complexity. Numerical simulation results demonstrate the superiorities of the proposed array geometry in resolving more sources than sensors, DOA estimation performance, and the angular resolution.

Image set compression has recently become an active research topic due to the explosion of digital photographs. In order to efficiently compress the image sets of similar images with moving objects, in this paper, we propose a novel algorithm for image set compression using multiple reference images. First, for an image set, its depth-constrained minimum arborescence is generated. We then present a reference image candidate determination method to build the reference image candidates for the images of the set. Furthermore, we propose a rate-distortion optimized multiple reference image selection method. This method compares the correlation between every image and each of its reference image candidates to produce its multiple reference images. Finally, compressed image data are achieved by employing block-based motion compensation and residue coding. In addition, we also give a new way of access to images to keep the same access delay with single reference image-based schemes. Compared with the state-of-the-art image compression algorithms, experimental results show that our proposed algorithm can significantly improve the image compression performance.

Negative selection algorithm (NSA) is an important method for generating detectors in artificial immune systems. Traditional NSAs randomly generate detectors in the whole feature space. However, with increasing dimensions, data samples aggregate in some specific subspaces, not uniformly distributed in the whole space. The detectors randomly generated by traditional NSAs cannot exactly fall into these specific subspaces, which results in a low coverage of detectors and a poor performance in a high-dimensional space. To overcome this defect, an improved real NSA based on subspace density seeking (SDS-RNSA) is proposed in this paper. In an SDS-RNSA, a subspace density seeking algorithm is adopted to procure the dense subspace regions of samples. Then, detectors are generated in each subspace region to cover up nonself-region efficiently and improve the performance of the algorithm. During the process of detector generation, the redundancy of candidate detectors is calculated, and the redundant is eliminated to minimize the time expense of the algorithm. Experimental results demonstrate that, compared with the classic NSAs, the SDS-RNSA can significantly improve the detection rate with an approximative false alarm rate and a smaller time expense. At the best case, the detection rate of the SDS-RNSA is increased by 14.7%, while the time expense is decreased by 78.1%.

We propose an effective framework for salient region detection in natural images based on the concept of self-guided statistical non-redundancy (SGNR). Salient regions are unique, because they have low information redundancy within a given image, while the rest of the scene may highly be redundant. We first analyze the structural characteristics of the image using structured image elements (samples) and classify them as being non-redundant or redundant based on textural compactness and overall non-redundancy. This guides saliency detection toward regions with low information redundancy by considering explicitly high information redundancy of samples potentially belonging to the background. We then compute the saliency map by determining the statistical non-redundancy of each sample using a conditional graph model. Experimental results based on publicly available data sets show that SGNR provides promising results when compared with existing saliency approaches.