Search

Abstract In this paper, an optimized complex nonnegative tensor factor 2D deconvolution (CNTF2D) is proposed to separate the sources that have been mixed in an underdetermined reverberant environment. Unlike conventional methods, the proposed model decomposition is performed directly on the statistics in the form of spectral covariance matrix instead of the data itself (i.e. the mixed signal). For faster convergence the model is adapted under the hybrid framework of the generalized expectation maximization and multiplicative update algorithms. This paper also proposes a solution to the issue of optimizing the model order i.e., number of components and convolutive parameters in the CNTF2D model. To this end, a latent-observation model based on Gamma-Exponential process is developed. In addition, the proposed Gamma-Exponential process can be used to initialize the parameterization of the CNTF2D. The proposed algorithm encodes a set of variable sparsity parameters derived from the Gibbs distribution. This permits a stable update and optimizes the CNTF2D with the correct degree of sparseness in the time-frequency domain. Experimental results on the underdetermined reverberant mixing environment have shown that the proposed algorithm is effective at separating the mixture with an average signal-to-distortion ratio of 2.5 dB. Highlights • Proposal of a complex tensor deconvolution model with full-rank spatial covariance matrix. • Tensor factorization works directly on the statistics (spectral covariance matrix) instead of the mixed signal. • Optimizing the sparsity level of each individual temporal code via the Gibbs distribution. • Optimizing the tensor model order based on Gamma-Exponential observation process. [ABSTRACT FROM AUTHOR]

This paper concentrates on the location methods for strictly noncircular sources by widely separated arrays. The conventional two-step methods extract measurement parameters and then, estimate the positions from them. Compared with the conventional two-step methods, direct position determination (DPD) is a promising technique, which locates transmitters directly from original sensor outputs without estimating intermediate parameters in a single step, and thus, improves the location accuracy and avoids the data association problem. However, existing DPD methods mainly focus on complex circular sources without considering noncircular signals, which can be exploited to enhance the localization accuracy. This paper proposes a maximum likelihood (ML)-based DPD algorithm for strictly noncircular sources whose waveforms are unknown. By exploiting the noncircularity of sources, we establish an ML-based function in time domain under the constraint on the waveforms of signals. A decoupled iterative method is developed to solve the prescribed ML estimator with a moderate complexity. In addition, we derive the deterministic Cramér–Rao Bound (CRB) for strictly noncircular sources, and prove that this CRB is upper bounded by the associated CRB for circular signals. Simulation results demonstrate that the proposed algorithm has a fast convergence rate, and outperforms the other location methods in a wide range of scenarios. [ABSTRACT FROM AUTHOR]

Abstract Extracting key frames from a video can reduce redundancies in continuous scenes and pithy represent the entire video. This technique copes with the issue of how to efficiently manage a large amount of video data and has many applications such as video indexing, querying, and browsing. Current key frame extraction methods often utilize sparse modeling, which assumes that each video frame can be expressed as a linear combination of a few representative key frames. Although these methods are successful, they are less aware of the fact that videos have spatio-temporal structure and used ℓ 1 norm based sparsity measure. In this paper, we propose a key frame extraction method that considers spatio-temporal information of videos. We employ a sparsity measure that is based on the determinant of the Gram matrix computed from entire video frames. The determinant sparsity measure is computed for the entire matrix, not for single frames, and thus it reflects the spatial or joint sparseness of the entire video. With this measure, the solutions are sparser than conventional sparseness measures but the cost function is non-convex and optimization becomes harder. Then, utilizing the fact that the proposed cost function is the difference of two convex functions, we employ an efficient algorithm based on the difference-of-convex (DC) programming, which often finds the global solution of the non-convex cost function. Experiments show that the proposed algorithm generates high-quality, high-compression key frames, compared with the existing key frame extraction method with the ℓ 1 norm. The determinant sparsity measure was recently proposed and can be utilized for more video processing applications such as video saliency detection. [ABSTRACT FROM AUTHOR]

Abstract This paper proposes the theoretical foundations of a novel radar detection method, based on analysis and decision making in the moments space of the scattered signal parameters. Through an adaptive process and from large-size samples of these parameters, several normal-distributed moments are calculated, which allow assigning each resolution cell of the searching region to the background or anomaly classes. Following the Neyman–Pearson criterion two optimal detection algorithms are proposed, one for statistically independent moments and other for correlated moments. Finally, three background-anomaly pairs are simulated, taking the amplitude of the video signal as parameter and two of its moments for decision making. As result, the detection curves for two sizes of the sample are presented, showing the possibilities of the proposed method. [ABSTRACT FROM AUTHOR]

Abstract In this paper, we propose a novel kernel adaptive filtering algorithm, which called variable learning rates kernel adaptive filter with single feedback (SF-VLRKAF). Based on a feedback structure, the past information can be used to estimate current output to improve the filtering performance, because of a momentum term existing in the weight update equation. A switch ON–OFF normalized variable learning rate is developed to obtain a tradeoff between convergence rate and filtering accuracy of the proposed algorithm. The weights, in the SF-VLRKAF, are updated at each iteration to avoid local minimum, and the analysis of mean square convergence is performed. Furthermore, a sufficient condition ensuring mean square convergence is obtained by applying the energy conservation relation. Moreover, we derive the lower and upper bounds on a theoretical result of the steady-state excess mean square error. Simulations for nonlinear regression, chaotic time-series predictions and real-world applications are presented to illustrate the effectiveness of the new algorithm. Graphical abstract Highlights • A KAF with single feedback and variable learning rates is designed. • The mean square convergence of proposed KAF is analyzed. • A sufficient condition for convergence of proposed KAF is obtained. • Simulation results show that the proposed KAF has superior performance. [ABSTRACT FROM AUTHOR]

Abstract In this paper, we present an efficient algorithm for identifying and tracking the impulse response of a sparse system that exhibits time varying sparseness and is driven by correlated input. The proposed method convexly combines the outputs of two filters, namely, the sparsity unaware affine projection algorithm (APA), and the sparsity aware zero attracting affine projection algorithm (ZA-APA), each trying to identify the same system using the same input. The combining parameter is adapted by following a steepest descent of the error variance at the convex combination output. A detailed performance analysis of the proposed combination is carried out, which reveals that while for highly non-sparse and highly sparse systems, the proposed combination converges respectively to the APA and ZA-APA (i.e., better of the two filters under the given levels of sparsity), for certain sparsity ranges, it leads to a combination filter that performs better than both the constituent filters. The claims made are validated by exhaustive simulation studies using white, colored as well as speech inputs. [ABSTRACT FROM AUTHOR]

The instantaneous frequency (IF) is a very important feature of nonstationary signals in numerous applications. The first overview of the concept and application of the IF estimators is presented in seminal papers by Boashash. Since then, a significant knowledge has been gained about the performance of the IF estimators. This knowledge has been used not only for development of various IF estimators but also for introduction of novel time–frequency (TF) representations. The IF estimation in environments characterized by low signal-to-noise (SNR) has achieved significant benefits from these theoretical developments. In this paper, we review some of the most important developments in the last two decades related to the concept of the IF, performance analysis of IF estimators, and development of IF estimators for low SNR environments. [ABSTRACT FROM AUTHOR]

In this paper, a blind image deblurring method is proposed using sparse representation with external patch priors. Different from traditional sparse-based methods that employ only internal priors from blurred images, additional external information is adopted to reconstruct latent images. In details, the Expected Patch Log Likelihood (EPLL) is introduced as a useful tool to describe external patch priors with a pre-trained Gaussian mixture model. With a set of operations, the EPLL is subsequently incorporated as a regularization term into the existing sparse-based deblurring model. Meanwhile, the dictionary is also carefully designed for each patch of the latent image, where atoms are obtained from the covariance matrix of the corresponding Gaussian component. A deblurring framework is further presented along with our sparse-based model. The solutions are respectively given to efficiently optimize the latent image and the blur kernel with an iterative procedure. The experiments demonstrate that our proposed algorithm achieves a competitive performance compared with the state-of-the-arts. Especially, it not only can obtain more accurate kernels for the deblurring, but also outperforms in noise reduction and artifact suppression for the restored images. [ABSTRACT FROM AUTHOR]

In this paper, a sparse reconstruction based approach to estimate Stokes parameters and 2-D DOAs of polarized sources is proposed in case of unknown number of sources. An L-shaped coprime array composed of dipole-dipole pairs, where a dipole in the dipole-dipole pair is placed on x-axis and the other is placed on z-axis, is used to reduce the mutual coupling across two dipoles in the dipole-dipole pair. In order to estimate the elevation of polarized sources, an orthogonal matching pursuit algorithm with the identification of the sources' number (OMP-isn) is proposed. In our proposed approach, the quasi deterministic maximum likelihood (qDML) test step is added in each iteration of algorithm. In addition, a refining grid step is also added in each iteration to reduce the estimation error incurred by the grid mismatch. The OMP algorithm and sparse iterative covariance-based estimation (SPICE) method are employed to estimate the azimuth of polarized sources. In order to pari the azimuth with the elevation, a pair-matching method is proposed based on the l 2 -norm and l 1 -norm ( l 2 / l 1 ) minimization. Based on the pairing azimuth and elevation, Stokes parameters of polarized sources are estimated by using Least Squares (LS) or Total Least Squares (TLS). The stochastic CRB is derived for the 2-D DOAs and Stokes parameters of polarized sources. Simulations are used to verify the performance of the proposed approach. [ABSTRACT FROM AUTHOR]

For wideband array signals, existing self-calibration methods use a first order approximation to simplify the estimation procedure under the assumption of small sensor position errors. However, the performance will degrade severely as the position errors increase. In this paper, a novel self-calibration method based on the time-frequency distributions (TFDs) is proposed to relieve the small errors constraint. We firstly develop an approach for wideband signals to select the TFDs of each source. Then the TFDs are utilized to obtain the propagation delays, which have a simple linear relationship with the sensor positions. The direction of arrival (DOA) and positions are finally estimated by a two-step iteration without any approximation. Simulation results demonstrate that the proposed method can achieve accurate DOA and position estimation over a broad range of position errors with lower computational complexity, and outperforms existing self-calibration algorithms. [ABSTRACT FROM AUTHOR]

For antenna array with large number of elements, the multistage Wiener filter (MWF) is an efficient method to perform reduced-rank beamforming. Similar to the generalized sidelobe canceller (GSC), MWF is also sensitive to direction of arrival (DOA) mismatch of the signal of interest (SOI). This condition limits the application of MWF to monopulse beamforming, in which the actual DOA of SOI may mismatch the assumed direction (i.e. the pattern look direction) due to target movement and platform disturbance. In this paper, we develop a variable-loaded multistage Wiener filter with spatial blocking broadening and automatic rank selection (VLSRS-MWF) to perform robust reduced-rank monopulse beamforming. Firstly, a null-broadened spatial blocking matrix which can be used for both the sum and difference beams is developed instead of the derivative constraints and the eigenvector constraints based blocking matrices. Secondly, a l 2 -norm regularization method is incorporated with MWF to further improve the output signal to noise ratio and to restrict the pattern sidelobe level. Then a Ritz value criterion with deterministic threshold for selecting the rank of the reduced-rank subspace is developed utilizing the eigenvalue distribution of sample covariance matrix and the Ritz value criterion. Simulation results show that the proposed VLSRS-MWF based monopulse algorithm can maintain good robustness and high angle estimation accuracy in presence of DOA mismatch. [ABSTRACT FROM AUTHOR]

Zero-Forcing (ZF) algorithm achieves the near-optimal detection performance for massive multiple-input multiple-output (MIMO) systems at expense of performing the complicated matrix inversion of a high dimensional matrix. In this paper, a novel Lanczos-algorithm-based signal detection method with soft-output is proposed to iteratively realize ZF algorithm for multiuser massive MIIMO systems, which avoids the exact computation of matrix inversion and in turn reduces the computational complexity from O ( K 3 ) to O ( K 2 ) , where K denotes the number of users. In the development of the proposed method, by analyzing the iterative process of Lanczos algorithm, an approximate low-complexity scheme is proposed to calculate the log likelihood ratios (LLRs) for soft channel decoding. Simulation results show that the proposed detector provides a relatively good tradeoff between the complexity and performance compared with the several recently proposed detectors, and achieves almost the same performance as the ZF algorithm with only 3 iterations. [ABSTRACT FROM AUTHOR]

An electronic support system receiver which is a passive receiver picks up an interleaved stream of pulses and extracts their pulse parameters. These parameters are sent to a deinterleaving subsystem which sorts them and forms pulse cells that each are assumed to belong to a specific emitter. In this paper, we develop a method for this task of deinterleaving of radar pulse sequences. For this aim, a novel pulse amplitude tracking algorithm is proposed for dynamically varying signal environments wherein radar parameters can change abruptly. This method particularly works for air-to-air engagements where pulse amplitude distortion due to channel effects can be considered negligible. Simulation results show that the proposed algorithm incorporated with a clustering algorithm improves deinterleaving of radar emitters that have agile pulse parameters such as airborne radars. [ABSTRACT FROM AUTHOR]

The sign algorithm with a fixed step-size is incapable of addressing the conflicting requirements between fast convergence speed and low steady-state misadjustments. In order to deal with this problem, a Rayleigh weighted gradient vector based variable step-size sign algorithm is proposed in this paper. In the new algorithm, the variable step-size is updated by the squared norm of a Rayleigh weighted sign gradient vector. The proposed algorithm can improve the convergence speed and tracking capability while maintaining the similar steady-state misadjustments in the presence of impulsive noises. A complex-valued energy conservation relation based convergence analysis is carried out to evaluate the convergence performance of the new algorithm. Simulation results are presented to verify the theoretical analysis and to demonstrate the desirable performance of the proposed algorithm. [ABSTRACT FROM AUTHOR]

Empirical wavelet transform (EWT) can successfully decompose a smooth or noise-free simulated signal into several components. This method has encountered difficulties in processing simulated signals containing noise and in applications. In order to improve the shortcomings of EWT and expand its application, this paper proposes a variable spectral segmentation EWT (VEWT) associated with the trend of spectral fluctuation. Different from scale-space representation and other optimized direct segmentation methods, this paper proposes a selective segmentation method. The extreme points of Multi-taper power spectral density (MPSD) are used to estimate the modes. An extended algorithm is proposed on the basis of Levenberg-Marquardt-Fletcher, which uses the extreme points and positions of MPSD to calculate the bandwidth corresponding to each mode. The information in each frequency band will be determined as the final mode. In the process of loop extraction, a set of boundaries associated with the fluctuations of the spectral will be obtained. The proposed method is more advantageous for the decomposition of signals containing noise. Simulated signals are used to verify the effectiveness of the proposed method. In addition, the MIT-BIH Arrhythmia Database is used to verify the applicability of the proposed method. [ABSTRACT FROM AUTHOR]

Multidimensional signal analysis has become an important part of many signal processing problems. This type of analysis allows to take advantage of different diversities of a signal in order to extract useful information. This paper focuses on the design and development of multidimensional data decomposition algorithms called Canonical Polyadic (CP) tensor decomposition, a powerful tool in a variety of real-world applications due to its uniqueness and ease of interpretation of its factor matrices. More precisely, it is desired to compute simultaneously the factor matrices involved in the CP decomposition of a real nonnegative tensor, under nonnegative constraints. For this purpose, two proximal algorithms are proposed, the Monotone Accelerated Proximal Gradient (M-APG) and the Non-monotone Accelerated Proximal Gradient (Nm-APG) algorithms. These algorithms are implemented via a regularization function with a simple control strategy capable of efficiently taking advantage of previous iterations. Simulation results demonstrate better performance of the two proposed algorithms in terms of accuracy when compared to other nonnegative CP algorithms in the literature. • All-at-once way approach for Nonnegative CP. • Accelerated Proximal Gradient algorithms for NCP Tensor Decomposition. • NCP decomposition under regularization constraints. • Avoiding collinear matrix factors problem with Monotone-APG and Non-Monotone-APG. [ABSTRACT FROM AUTHOR]

In this paper, a novel localization algorithm for mixed far-field and near-field sources is proposed in the presence of unknown mutual coupling. Based on the principle of rank reduction, direction-of-arrival (DOA) estimates of far-field sources are firstly decoupled under unknown mutual coupling. Then these estimates are employed to generate the mutual coupling coefficients. Finally, by the mutual coupling compensation and the far-field components elimination, near-field sources parameters (DOA and range) are obtained. The proposed algorithm is efficient in that it only requires second order statistics and one dimensional spectral search. Simulation results demonstrate that our algorithm is effective for the classification and localization of mixed sources under unknown mutual coupling. [ABSTRACT FROM AUTHOR]

This paper proposes a novel algorithm for the voltage fluctuation detection of a power system based on sparse representation modeling. The contents of this research mainly include: (1) By first constructing a proper objective function of the frequency and phase, we convert the fundamental signal estimation problem to a simple mathematical convex optimization problem, which can be easily solved using an exhaustive search strategy. (2) From the viewpoint of signal restoration, we regard the voltage fluctuation detection as a signal inpainting problem and then develop an l 0 norm-based optimization equation that exploits the sparsity prior of fluctuation component to recover the desired representation vector. (3) With the assumption that the voltage envelope changes smoothly, we establish an l 2 norm-based regularization equation to further improve the regularity of the result. Experimental results show that the proposed algorithm performs well on demodulating the fundamental signal and voltage fluctuation component, with good ability of noise robustness, when compared to the classical Hilbert transform-based detection method and square demodulation method. [ABSTRACT FROM AUTHOR]

This paper presents a method for improving users' quality of experience through processing of movie soundtracks. The dialogue clarity enhancement algorithms were introduced for detecting dialogue in movie soundtrack mixes and then for amplifying the dialogue components. The front channel signals (left, right, center) are analyzed in the frequency domain. The selected partials in the center channel signal, which yield high disparity between left and right channels, are detected as dialogue. Subsequently, the dialogue frequency components are boosted to achieve an increased dialogue intelligibility. Techniques for reduction of artifacts in the processed signal are also introduced. It is done through smoothing in the time domain and in the frequency domain, applied to reduce unpleasant artifacts. The results of objective and subjective tests are provided, which prove that an increased dialogue intelligibility is achieved with the aid of the proposed algorithm. The algorithm is particularly applicable in mobile devices while listening in changing conditions and in the presence of noise. [ABSTRACT FROM AUTHOR]

The aim of this paper is to improve the energy concentration of the Stockwell transform (S-transform) in the time–frequency domain. A modified S-transform is proposed with several parameters to control the width of a hybrid Gaussian window. A constrained optimization problem is proposed based on an energy concentration measure as objective function and inequalities constraints to define the bounds of the Gaussian window. An active-set algorithm is applied to resolve the optimization problem. The optimization of the energy concentration in the time–frequency plane can lead to more reliable applications for non-stationary signals. The simulation results show a significant improvement of the proposed methodology most notably in the presence of noise comparing with the standard S-transform and existing modified S-transform in the literature. Moreover, comparison with other known time–frequency transforms such as Short-time Fourier transform (STFT) and smoothed-pseudo Wigner–Ville distribution (SPWVD) is also performed and discussed. The proposed S-transform is tested also on real non-stationary signals through an example of split detection in heart sounds. [ABSTRACT FROM AUTHOR]

Several techniques have been proposed in the literature to accelerate the convergence of adaptive algorithms for the identification of sparse impulse responses (i.e., with energy concentrated in a few coefficients). Among these techniques, the improved µ-law proportionate normalized least mean squares (IMPNLMS) algorithm is one of the most effective. This paper presents an accurate transient analysis of this algorithm and derives an estimate of its steady-state MSE, without requiring the assumption of white Gaussian input signals. [ABSTRACT FROM AUTHOR]

A key challenge in the field of graph signal processing is to design proper transform methods to extract valuable information from signals on weighted graphs. This paper first defines convolution, translation and modulation operators in graph fractional domain, and proposes windowed graph fractional Fourier transform (WGFRFT). Related properties are proved. Then, a fast algorithm of WGFRFT is designed to increase flexibility. The robustness and superiority of the fast algorithm are verified via simulations with synthetic signals on graphs. Finally, an application to graph clustering is presented to show the practicability of WGFRFT. [ABSTRACT FROM AUTHOR]

Model-based adaptive signal decomposition algorithm is an important research topic in the field of signal analysis. The amplitude modulated and frequency modulated (AM-FM) model is the most common model in the signal. In this paper, an adaptive signal decomposition algorithm based on the complex AM-FM model is proposed. This algorithm first obtains the analytical expression of the analyzed signal; and then the optimal equation of the analytic signal based on the model is realized by initializing the instantaneous frequency (IF) of the model. Finally, the partial differential equation and the alternating direction method of multipliers technique are used to solve the optimization equation to achieve a high precision estimation of the model. Experiments on simulated and real-life data verify that the proposed algorithm can accurately estimate the instantaneous amplitude (IA) and IF of each component in the signal, and has strong robustness and high efficiency. [ABSTRACT FROM AUTHOR]

Direction-of-arrival (DOA) estimation is a fundamental problem in many applications of array signal processing. In practice, it is more important to track the DOAs of moving target signals. In this paper, we study the problem of tracking the DOAs of multiple signals in the frequency diversity array (FDA)-MIMO radar system. Aiming at FDA signals, we propose an efficient algorithm for DOA tracking, which combines subspace updating with the DOA estimation of the FDA signal. The proposed algorithm effectively solves the problem of angle-range coupling in the FDA signal. Moreover, it achieves excellent tracking performance and requires low computational complexity. Simulation results demonstrate the effectiveness of the proposed algorithm. [ABSTRACT FROM AUTHOR]

In this paper, a novel and complete derivation of the range doppler algorithm (RDA) formulation is developed for stepped frequency continuous wave (SFCW) synthetic aperture radar (SAR) usage. As the RDA has been originally developed for linear frequency modulated (LFM) pulse excited SAR data, the detailed mathematical derivation of SFCW based RDA formulation is presented with similarities, differences and additions to the traditional LFM-RDA formulation for comparison. Then, the developed SFCW-RDA was tested with simulation data and the performance of the algorithm is determined by examining some important parameters including impulse response width, peak to sidelobe ratio, integrated side lobe ratio, range resolution, and azimuth resolution. The performance metrics of the SFCW-RDA in comparison to LFM-RDA thru the constructed SAR images have proven the implementation validity of the proposed technique. [ABSTRACT FROM AUTHOR]

In the graph signal processing literature, most methods were developed based on the assumption of Gaussian noise since it can lead to computationally efficient and mathematically tractable solutions. Unfortunately, the Gaussian distribution cannot capture the sharp spikes and tail heaviness of the signal noise in various natural phenomena. The α -stable distribution is a generalization of the Gaussian distribution and is a more appropriate model for such impulsive heavy-tailed noise. In this paper, we consider the problem of adaptive estimation and sparse sampling for signals defined over graphs in the presence of impulsive α -stable noise. To tackle the problems associated with α -stable noise, the graph signal estimation problem is formulated as a minimum dispersion (MD)-based optimization. A novel adaptive least mean p th power (LMP) algorithm is proposed for robust estimation of band-limited graph signals from partial observations in α -stable noise environments. The mean square performance of the proposed LMP algorithm is theoretically analyzed. To handle the case of band-limited graph signals with unknown and time-varying bandwidth and spectral contents, the sparse nature of the Graph Fourier transform of the signal is exploited to develop an adaptive graph sampling technique. Specifically, the proposed sparse sampling technique identifies the spectral signal support via sparse online estimation building on the iterative shrinkage-thresholding algorithm. Numerical simulation studies are presented to corroborate the performance advantages of the proposed adaptive estimation and sparse sampling algorithms. [ABSTRACT FROM AUTHOR]

Over the last decade remote photoplethysmography (rPPG) algorithms have been developed extensively. As a result, pulse rate can now be accurately estimated by video data for still subjects. However, in realistic conditions both the accuracy of these algorithms and benchmark datasets are far from perfect. In this paper we propose a new rPPG method which enables heart rate detection by video from a standard webcam. The algorithm is robust with respect to such factors as illumination changes or the subject's movements, and can track fast pulse rate changes. To do that, the algorithm determines the approximate value of the pulse rate and then specifies it with high time resolution. In order to comprehensively study the proposed method, we collected a new dataset consisting of videos recorded in various challenging conditions of several categories as well as reference photoplethysmograms recorded synchronously with a contact pulse oximeter. The proposed method showed high performance under all conditions including blinking illumination, speech and large-amplitude movements. We tested two simplified versions of the algorithm, which provided competitive scores as well. However, with human movement videos the full method showed better results than its simplified versions (p<0.001). The proposed algorithm was tested on the existing UBFC-RPPG database and compared with previous methods. Our method showed high results (2.10 MAE, 3.43 RMSE). • A novel robust method for remote photoplethysmography signal processing is proposed. • Pulse rate is approximately estimated and then refined in time and frequency domains. • A new publicly available challenging rPPG dataset called MoLi-ppg is introduced. [ABSTRACT FROM AUTHOR]