Search

In this paper, we explore efficient hardware implementation of feedforward artificial neural networks (ANNs) using approximate adders and multipliers. Due to a large area requirement in a parallel architecture, the ANNs are implemented under the time-multiplexed architecture where computing resources are re-used in the multiply accumulate (MAC) blocks. The efficient hardware implementation of ANNs is realized by replacing the exact adders and multipliers in the MAC blocks by the approximate ones taking into account the hardware accuracy. Additionally, an algorithm to determine the approximate level of multipliers and adders due to the expected accuracy is proposed. As an application, the MNIST and SVHN databases are considered. To examine the efficiency of the proposed method, various architectures and structures of ANNs are realized. Experimental results show that the ANNs designed using the proposed approximate multiplier have a smaller area and consume less energy than those designed using previously proposed prominent approximate multipliers. It is also observed that the use of both approximate adders and multipliers yields, respectively, up to 50% and 10% reduction in energy consumption and area of the ANN design with a small deviation or better hardware accuracy when compared to the exact adders and multipliers. [ABSTRACT FROM AUTHOR]

In the last 20 years, most of the human demise have been due to improper functioning of the heart. ECG is a very essential, initial clinical test for diagnosing dangerous cardiac electrical disturbances. Many techniques have been proposed to analyze and process ECG signals. It is difficult to check the performance conspicuously because the actual ECG is corrupted by several sources of noise and artifacts. In view of this, the generation of an artificial ECG signal that can reflect all the characteristics of a real ECG signal is a very challenging task in biomedical signal processing. Classification of ECG data aids early detection and prediction of any heart-related disease so that timely treatment can be suggested by the clinicians. This paper presents a method of modeling an ECG signal based on the parametric cubic splines and derives new data set therefrom, followed by ECG classification, with three different classification techniques with the aid of Orange software. The support vector machine (SVM), CN2 rule Induction and tree classifiers have been tested on the model and the Tree algorithm has shown the best performance metrics. This approach is applied for normal and abnormal sinus rhythms. Analysis and processing of these modeled ECG signals eliminate the requirement of pre-processing and conventional feature extraction techniques. [ABSTRACT FROM AUTHOR]

In this paper, the problem of single-channel blind source separation (SCBSS) is addressed using a novel approach that combines the adaptive mode separation-based wavelet transform (AMSWT) and the density-based clustering with sparse reconstruction. The proposed method is performed in the time–frequency domain and in a reverberant environment. First, using the Fourier transform, the amplitude spectrum of the observed mixture signal is obtained. Then, using variational scaling and wavelet functions, the AMSWT is used to adaptively extract spectral intrinsic components (SICs). To obtain a better time–frequency resolution, the AMSWT is applied to each mode. Thus, the SCBSS problem is transformed into a non-underdetermined. Then, for each frequency bin, the density-based clustering, reformulated to the eigenvector clustering problem, is performed to estimate the mixing matrix. Finally, sparse reconstruction is used to reconstruct the sources. The proposed method is evaluated using objective measures of separation quality. A computational complexity evaluation based on time consumption is also performed. Simulation results show that the proposed method is very effective for solving the SCBSS problem and provides better separation performances than the reference methods. However, the proposed method is computationally expensive. [ABSTRACT FROM AUTHOR]

The estimation of the Ideal Binary Mask (IBM) based on speech cochleagram and visual cues were carried out in this paper to improve the speech intelligibility and quality using an Audio-Visual Convolutional Neural Network (AVCNN). Many speech enhancement techniques in the past depended heavily on audio attributes to reduce the noise present in the speech signal. Several studies have recently revealed that speech enhancement using visual data as an auxiliary input with audio data is more effective in reducing acoustic noise in speech signals. In the proposed work the multichannel CNN is used to extract the dynamics of both visual and audio signal features which were then integrated to estimate the threshold using the proposed algorithm to obtain the IBM for the enhancement of speech signal. The performance of the proposed model is evaluated primarily to measure the speech intelligibility in terms of STOI, ESTOI, and CSII additionally speech quality is also measured in terms PESQ, SSNR, CSIG, CBAK, and COVL. The evaluation results reveal that the proposed audio-visual mask estimation model outperforms the Audio-only, Visual-only, and existing audio-visual mask estimation models. The proposed AVCNN model, in turn, demonstrates its efficiency in merging the dynamics of audio information with visual speech information for speech enhancement. [ABSTRACT FROM AUTHOR]

Superpixel segmentation has been widely used in computer vision as a preprocessing to reduce the number of image primitives for subsequent image processing tasks. Simple Linear Iterative Clustering (SLIC) is the most popular superpixel segmentation algorithm in the literature, which is an iterative algorithm. The basic SLIC technique includes two main steps named Assignment and Update steps, and their data dependency is the main problem to speed up execution of the basic SLIC algorithm. It is also well known that this algorithm is not memory efficient, which is the main bottleneck in applying basic SLIC in low-power applications. In this paper, we have proposed a new structure for SLIC implementation. This new structure replaces the cluster-based search operation with pixel-based process in the Assignment step that gains considerable speed improvement. This new technique simply removes the data dependency in the SLIC structure. In the proposed method, the distance index, the main parameter computed in the Assignment step, no longer is needed to be stored in memory. This leads to a significant improvement in the memory efficiency of the proposed structure. The proposed structure is implemented on the Virtex-5 FPGA family, and the results show about 26 % in memory improvement compared to basic SLIC. According to the implementation results, the proposed technique process 143 frames per second for an input image of 300 × 400 pixel resolution which is far faster than real-time process requirements. [ABSTRACT FROM AUTHOR]

Coherently Distributed (CD) sources are suitable for practical direction-of-arrival (DOA) estimation in massive Multiple-Input Multiple-Output (MIMO) communication systems. However, massive MIMO systems employ a large number of antennas to improve performance, which significantly raises the complexity of conventional DOA estimation algorithms for CD sources. To reduce the complexity, a novel low-complexity CD unitary estimation of signal parameters via the rotational invariance technique (U-ESPRIT) algorithm based on discrete cosine transformation (DCT) beamspace transformation and multistage Wiener filter (BU-ESPRIT) is proposed in this paper. To concentrate the signal energy and make better use of beamspace features, the DCT transformation is used to extract the signal. For the sake of reducing the complexity and achieving accurate angle estimation, U-ESPRIT is performed in beamspace. Moreover, the covariance matrix matching process and eigenvalue decomposition of the covariance matrix are substituted by forward recursion of the multistage Wiener filter to further reduce the complexity of the proposed algorithm. Theoretical analysis indicates the computational superiority of the proposed algorithm over that of some existing algorithms. Simulation results demonstrate that the proposed algorithm can be utilized to efficiently estimate angles of CD sources in massive MIMO systems. [ABSTRACT FROM AUTHOR]

Digital filter design optimization counts as one of the most challenging problems in circuit development. Furthermore, the bit-width allocation has a crucial impact on the efficiency and accuracy of fixed-point digital filters. The conventional implementation procedure involves studying the bit-width allocation problem after determining the filter coefficients. This sequential procedure leads frequently to an excessive bit-width allocation, hence an extra implementation cost. The main contribution of this paper is to propose a new implementation procedure based on the simultaneous handling of both design issues and the bit-width optimization for the infinite impulse response filter. In this study, the problem is formalized by a multi-objective programming technique using the non-dominated sorting genetic algorithm II. The joint consideration of the two problems offers a better trade-off between design and bit-width allocation. The output of our approach is compared to the results of existing algorithms. Simulation results show that the multi-objective method significantly reduces the implementation cost while ensuring filter stability and an acceptable mean square error (MSE) compared to the existing algorithms. [ABSTRACT FROM AUTHOR]

In this paper, we develop a robust direction-of-arrival (DOA) estimation algorithm with a nested array with unknown mutual coupling. By using the matrix transformation of the product of mutual coupling matrix and steering vector, we firstly derive a coarray signal model including a manifold matrix without mutual coupling effect. Subsequently, we build the block sparse representation of the coarray signal by exploiting the sparsity of the signals. Finally, we estimate the DOAs of sources by formulating a simplified block sparse recovery problem. The proposed algorithm utilizes all coarray outputs and reduces the influence of mutual coupling effect, and thus can resolve more sources than the number of sensors. Numerical results demonstrate the superiority of the proposed algorithm over several existing techniques. [ABSTRACT FROM AUTHOR]

Speckle noises widely exist in ultrasound images. They seriously affect the quality of images and cause the doctor to make mistakes in diagnosis. In this paper, we propose a three-path parallel convolutional neural network called USNet to achieve speckle reduction for ultrasound images. We combine three different sub-networks to increase the width of the whole network instead of the depth. The ideas of dilated convolution and shortcut connection are added to increasing the learning ability of the model. These make our proposed USNet can learn deeper information from the original ultrasound image. In experiments, we verify the effectiveness of the proposed three-path parallel structure and the dilated convolution by conducting ablation experiments. At the same time, we propose a different method to construct the training dataset. For the noisy training image, we artificially add speckle noise at three different σ levels to enhance the generalization performance of the proposed method. For the noise-free true labels, we use two stages to obtain on the basis of original images, including Optimized Bayesian Non-Local Means with block selection method (OBNLM) and Second-order Oriented Partial-differential Equation (SOOPDE) method. Then we compare our proposed method with other four different methods, including Kuan method, Speckle Reducing Anisotropic Diffusion (SRAD) method, the OBNLM method, and Residual Learning Network (RLNet). We qualitatively and quantitatively evaluate these methods in terms of smoothness, texture information protection, and edge clarity. The results show that our proposed USNet model can batch and quickly achieve good speckle reduction as well as texture preservation for different types of ultrasound images without any parameter adjustment. The USNet model has the advantages of good adaptability, robustness, and generalization. It is of great significance for improving the diagnostic efficiency of clinical medicine. [ABSTRACT FROM AUTHOR]

This paper deals with the state feedback control of singular continuous-time linear systems with discrete-time measurement of the state. Intuitively, descriptor or singular description of linear systems is more general than conventional state space description. In particular, a descriptor form includes information about algebraic as well as dynamic constraints. The use of an event-triggering mechanism results in a varying sampling-time period. Moreover, it is worth noting that the system is really in open loop during the inter-event time interval. The consequence is that there is no means to make the system admissible and by the way all the nice theory of admissibilization becomes useless. Hence, a continuous-time singular system in a sampled data closed-loop framework is badly posed unless the open loop is regular. In the case the open-loop system is regular, it is shown that the fast component of the state vector exhibits a discontinuous behavior. The dynamic event-triggering mechanism is shown to insure that the inter-event time interval is lower bounded, and as a consequence, it avoids the Zeno phenomenon. Simulations are presented to illustrate the obtained results. [ABSTRACT FROM AUTHOR]

Positive feedback source-coupled logic (PFSCL) circuits are imperative in high-resolution mixed-signal applications. In this paper, memristors are introduced in PFSCL style to amalgamate their advantages in circuit design. Two new memristor-based PFSCL architectures are presented. The first proposed architecture (Mem-PA1) employs memristors with a PFSCL inverter, while the other uses them with PFSCL NOR/OR gate (Mem-PA2). The proposed circuits eliminate the restriction of implementing the logic function in NOR/OR form or sum of two minterms in existing architectures. The reduced gate count gives significant advantage in power and propagation delay values. Different functions are implemented and simulated using CMOS PTM 180 nm technology parameters and a TiO2 memristor. A maximum improvement of 33%, 80% and 86% in propagation delay, power and power delay product, respectively, is observed for Mem-PA1 exclusive-OR(XOR3) gate with respect to existing architectures-based gates. Further, it is observed that Mem-PA2 XOR3 gate shows 5.16% improvement in delay over the Mem-PA1 gate for same value of power consumption. Lastly, the performance is also examined with technology scaling and at various process corners. [ABSTRACT FROM AUTHOR]

This paper studies the problem of identifying discrete-time Wiener systems consisting of polynomial nonlinearities and linear subsystems with unknown orders. The best linear approximation for a Wiener system is found by using finite generalized rational orthogonal basis functions. The novelty of this work is that the poles of the basis functions are adaptively selected by applying the proposed method. This selection leads to a greedy type of best linear approximation with a fast convergence rate. Further, the nonlinearity is determined by solving the conventional least-squares problem as usual. Moreover, the case in which there are errors in the frequencies is analyzed. The analytical results show that small perturbations have limited effects on the upper bounds of the estimation error. Two numerical examples are given to verify the validity of the proposed method. [ABSTRACT FROM AUTHOR]

In this paper, the medical image watermarking technique is performed by using lifting wavelet transform, singular value decomposition, and Hessenberg decomposition. Multimedia security is one of the most significant concerns in the current digital era even though it is effortless to reproduce, disseminate, and modify multimedia data. Digital watermarking is a technique for shielding critical information to protect it from illegitimate duplication and distribution. The performance of the proposed scheme is checked with several wavelets. For this purpose, various wavelets such as Haar, Daubechies, Symlet, biorthogonal, and reverse biorthogonal are applied. The scaling factor plays a significant role in watermarking. So, to get the optimized scaling factor, particle swarm optimization and JAYA optimization algorithms are used. The authentication process is done by using advanced encryption standards. The performance of the proposed algorithm is checked by applying different types of attacks such as salt and pepper noise, Gaussian noise, filtering, geometrical, and compression attacks. Peak signal-to-noise ratio, normalized correlation coefficient, mean square error, and structural similarity index measurement are calculated to evaluate the resilience and indistinguishability of the proposed watermarking algorithm. The results of the proposed technique show the significant improvement in the performance and can be used to protect the critical medical data. [ABSTRACT FROM AUTHOR]

This paper presents a robust distributed adaptive algorithm called diffusion recursive least M-estimate (DRLM), which enhances the robustness of the diffusion recursive least square (DRLS) algorithm against impulsive noise by incorporating the strong resistance to impulsive noise of the modified Huber function (MHF). However, the tracking speed of the proposed DRLM algorithm is not favorable in scenario where the network parameter vector of interest changes abruptly. To overcome this drawback, a new variable forgetting factor (VFF) strategy is devised and the VFF-DRLM algorithm is obtained. Then, the asymptotic unbiasedness and steady-state network mean square deviation (NMSD) performance of the DRLM algorithm are analyzed. Finally, the accuracy of the steady-state analysis result of the DRLM algorithm and the superiority of the proposed algorithms compared to other competing algorithms are demonstrated by extensive numerical simulations. [ABSTRACT FROM AUTHOR]

The presence of RF components in mixed-signal circuits make it a challenging task to resolve tradeoffs among performance specifications. In order to ease the process of circuit design, these tradeoffs are being analyzed using multi-objective optimization methodologies. This paper presents a hybrid multi-objective optimization framework (MHPSO), a combination of particle swarm optimization and simulated annealing. The framework emphasizes on preserving nondominated solutions in an external archive. The multi-dimensional space excluding the archive is divided into several sub-spaces according to a velocity-temperature mapping scheme. Further, the solutions in each sub-space are optimized using simulated annealing for generation of a Pareto front. The framework is extended by incorporating crowding distance comparison operator (MHPSO-CD) to maintain nondominated solutions in the archive. The effectiveness of proposed methodologies is demonstrated for performance space exploration of three electronic circuits, i.e., a two-stage operational amplifier, a folded cascode operational amplifier, and a low noise amplifier with inductive source degeneration. Additionally, the performance of proposed algorithms (MHPSO, MHPSO-CD) are evaluated on various test functions, and the results are compared with standard multi-objective evolutionary algorithms. [ABSTRACT FROM AUTHOR]

Due to reset process of phase frequency detector (PFD), the induced blind zone degrades the acquisition performance of phase-locked loop. Most of reported blind-zone free solutions are based on transistors, which cannot be implemented with field programmable gate array (FPGA). Therefore, a blind zone-suppressed PFD suit for FPGA is proposed in this paper. If any trigger condition of blind zone is met, the proposed reset mask or edge recovery operation is active. Thereafter, the corresponding edge of detected clock is kept or recovered to alleviate the effect of blind zone. At last, the proposed PFD is written by Verilog hardware description language, and implemented with Xilinx Virtex-7 XC7VX330TFFV1157-3. The occupied resources are 9 LUTs and 15 FFs, and the logic delay of critical path is only 0.264ns. The correctness of blind zone-suppressed PFD is also verified by simulation results. [ABSTRACT FROM AUTHOR]

In this paper, the minimum variance distortionless response (MVDR) beamforming technique is studied in the presence of a spatially coherently distributed (CD) source. In the first part, we propose the CD-MVDR beamforming in which the steering vector of the CD source model is used instead of the conventional point source model. We derive a theoretical expression of the white noise gain of CD-MVDR beamforming, which is inversely proportional to the square of the difference between the angular dispersion of the actual source and that of the CD-MVDR beamforming model. In the second part, based on the performance analyses, we propose an efficient optimization method for the microphone array geometry to reduce the impact of the CD source angular dispersion on the performance of the conventional MVDR beamforming. Simulation results validate our proposed theoretical expressions and show that with the proposed geometry, the conventional MVDR beamforming tends to be significantly more robust to the angular dispersion of the CD source for both white noise gain (WNG) and directivity factor (DF). [ABSTRACT FROM AUTHOR]

This paper is concerned with asynchronous finite-time H ∞ control for a class of discrete-time switched linear systems via admissible edge-dependent average dwell time (AED-ADT) approach. Firstly, by considering the switching time delay between the system and the state feedback controller, appropriate Lyapunov functions are constructed for asynchronous and synchronous switching, respectively. Secondly, for the existence of a set of state feedback controllers, a sufficient condition which guarantees the finite-time boundedness of the closed-loop system with AED-ADT is proposed. Thirdly, a sufficient condition for finite-time H ∞ control with a prescribed H ∞ performance is further developed based on the obtained result. Finally, a numerical example is given to verify the validity of the proposed theoretical results. [ABSTRACT FROM AUTHOR]

The dictionary grid mismatch problem in sparse processing is an important factor affecting parameter reconstruction. The existing grid correction algorithms are limited by the initial grid division and the accurate estimation of the support set vector. Furthermore, these methods fail if multiple scattering points fall into the same grid interval. In this paper, a constrained inhomogeneous grid optimization sparse processing algorithm is proposed. The proposed method includes two sub-processes, namely initial grid optimization and the support set vector correction. The optimization of the initial grid is to use the idea of the coordinate descent method to iteratively generate an updated grid vector, which is used for the inhomogeneous fission of the grid around the target from coarse to dense. Moreover, under the constraint of dictionary atom maximum correlation, multiple scattering points are separated in different grids, which improves the search accuracy of support set atoms. The correction process of the support set vector is to use the first-order Taylor expansion of the dictionary grid to linearly approximate the real parameters of the target, to achieve adaptive correction of dictionary mismatched atoms. According to the simulation experiment of the frequency agile radar scenario, the proposed algorithm can achieve higher range-Doppler parameter joint estimation accuracy in the multi-scattering point scenario in comparison with the conventional sparse recovery algorithm and grid point correction algorithm. [ABSTRACT FROM AUTHOR]

Artificial neural networks (ANNs) have gained considerable interest in industrial and academic research due to their vast applicability areas; consequently, many real-time applications using ANNs have been developed. In this paper, we propose an efficient multiplier-less realization architecture using distributive arithmetic (DA) for the perceptrons to realize MLP inference models. An efficient organization of the activation function unit using a domain-specific approach is proposed to save a significant number of resources. Further, a generalized ROM-based efficient realization architecture using DA is proposed for the calculation of the weighted sum. The architecture is further optimized for the area by exploiting the symmetry properties. Several previously reported MLP reference models are used for the purpose of comparison. Finally, with the help of application-specific integrated circuit synthesis results, we demonstrate that our proposed method delivers much better performance parameters. [ABSTRACT FROM AUTHOR]

Low-rank matrix recovery (LRMR) has a wide range of applications in computer vision. In recent years, the rank residual constraint (RRC) model, which aims to approximate the underlying low-rank matrix via minimizing the rank residual, has provided a new idea for LRMR. Different from the RRC model which employs l 1 norm as the regularization term, this paper exploits the weighted Schatten p-norm as the regularizer for the rank residual to obtain a new rank minimization model, namely the rank residual constraint model with adaptive weighted Schatten p-norm (SRRC). The proposed SRRC considers the importance of different rank residual components and gives better approximation to the original low-rank assumption. In the SRRC model, the adaptive p-values can vary with the sparsity of the rank residuals and different weights are assigned to different rank residuals. Thus, the proposed SRRC not only improves the sparsity of the rank residuals but also alleviates the over-penalty of the large singular values. We analyze the solution of SRRC model and prove that its global optimum can be efficiently solved by the generalized soft thresholding algorithm (GST). Extensive experimental results demonstrate that the proposed algorithm achieves favorable performance compared to many popular or state-of-the-art image denoising methods in terms of both objective and visual perception. [ABSTRACT FROM AUTHOR]

This paper proposes a diffuseness estimation-based single-source time–frequency point (SSTP) detection method for multisource direction of arrival (DOA) estimation. According to the composition, time–frequency (TF) points are divided into three types: SSTP, multisource TF, and interference TF. SSTPs and multisource TF points are defined as weak interference time–frequency points (WITPs). An SSTP is a TF point consisting only of the direct component of one sound source, which is beneficial for DOA estimation. Therefore, multisource DOA estimation is transformed into single-source DOA estimation by SSTP detection. Diffuseness estimation is introduced for a sound field microphone array. WITPs are detected by a diffuseness estimation–based detection method. Phase similarity determination is adopted to identify SSTPs from detected WITPs. Multiple sound source localization is completed by searching peaks in the normalized histogram of DOA estimates corresponding to the detected SSTPs. Experiments demonstrate that the proposed method achieves the precise detection of SSTPs, and evaluations show that it has superior accuracy of multiple sound source counting and localization in reverberant and noisy environments. [ABSTRACT FROM AUTHOR]

This paper presents a novel hybrid model reduction method to simplify large-scale continuous-time single-input–single-output and multi-input–multi-output dynamic systems using the advantages of the balanced truncation method and the particle swarm optimization (PSO) algorithm. The balanced truncation method obtains the reduced model denominator coefficients to ensure stability. The PSO algorithm minimizes the integral square error between the step responses of the original system and the reduced model as much as possible. It leads to the optimal numerator coefficients. The advantage of the proposed approach is that, for optimizing reduced model numerator coefficients, the search space boundaries of the PSO algorithm are not entirely random. They are selected using the balanced truncated reduced model numerator. So, the suggested method avoids two significant problems with evolutionary algorithms: the arbitrary choice of search space and the longer simulation time. Four power system models and four numerical examples from the literature are considered to assess the effectiveness of the proposed reduction method. The step responses and Bode diagrams for the higher-order system and its corresponding reduced models are displayed. For comparison, statistics are tabulated based on the rise time, the settling time, the peak overshoot, the integral square error, and the root-mean-square error. The proposed method ensures stability and other features of the higher-order system in the reduced model. The performance measure values and the time domain characteristics demonstrate the ability and effectiveness of the proposed approach. All the case studies use a MATLAB simulation environment. [ABSTRACT FROM AUTHOR]

In this paper, we demonstrate the significance of restoring harmonics of the fundamental frequency (pitch) in the deep neural network (DNN)-based speech enhancement. The parameters of the DNN can be estimated by minimizing the mask loss, but it does not restore the pitch harmonics, especially at higher frequencies. In this paper, we propose to restore the pitch harmonics in the spectral domain by minimizing cepstral loss around the pitch peak. Restoring the cepstral pitch peak, in turn, helps in restoring the pitch harmonics in the enhanced spectrum. The proposed cepstral pitch-peak loss acts as an adaptive comb filter on voiced segments and emphasizes the pitch harmonics in the speech spectrum. The network parameters are estimated using a combination of mask loss and cepstral pitch-peak loss. We show that this combination offers the complementary advantages of enhancing both the voiced and unvoiced regions. The DNN-based methods primarily rely on the network architecture, and hence, the prediction accuracy improves with the increasing complexity of the architecture. The lower complex models are essential for real-time processing systems. In this work, we propose a compact model using a sliding-window attention network (SWAN). The SWAN is trained to regress the spectral magnitude mask (SMM) from the noisy speech signal. Our experimental results demonstrate that the proposed approach achieves comparable performance with the state-of-the-art noncausal and causal speech enhancement methods with much lesser computational complexity. Our three-layered noncausal SWAN achieves 2.99 PESQ on the Valentini database with only 10 9 floating-point operations (FLOPs). [ABSTRACT FROM AUTHOR]

This paper focuses on the quasi-stabilization of the quaternion-valued fractional-order memristive neural networks. Based on the contraction mapping theory, a sufficient condition is derived to ensure the existence of the equilibrium point for the memristive neural networks. Subsequently, by means of Lyapunov functional and fractional Laplace transform, a algebraic inequality-based condition is developed to guarantee the quasi-stability of the equilibrium point. In addition, a related question is whether the convex closure proposed by the quaternion parameters is meaningful, to overcome this issues, a vector ordering approach is proposed, which can be used to compare the "magnitude" of two different quaternions. Finally, the corresponding simulation results are included to show the effectiveness of the proposed methodology derived in this paper. [ABSTRACT FROM AUTHOR]

The paper pioneers a thorough mathematical approach for the Lazzaro variant of the W(inner) T(ake) A(ll) maximum rank and amplitude analog selector. Two exact levels of output which split the maximum and determine the resolution, are found for the first time. At the input, a list of currents I 1 , I 2 , ... , I N from a large family L with smallest relative distance Δ on a 0 , I M scale is applied. To distinguish the largest current I w (the winner) from the second largest I l (the loser), the paper proposes two decision levels, D ¯ and D ̲ , for the output voltage list U 1 , U 2 , ... , U N . The upper level D ¯ is surpassed only by the U w winner and encodes the winning rank w. All other ranks are placed under the lower level D ̲ . Two rigorously treated optimization problems with inequality constraints lead to the identification of two input lists that yield the levels D ¯ and D ̲ as outputs. They are valid for processing any list in the L family. The index D ¯ - D ̲ / U M —"the output resolution"—expresses how large the gap between the first and the second component on the 0 , U M scale is. It exceeds "the input resolution," i.e., the similar index Δ / I M at the input and the two depend monotonically on each other. Widely commented numerical examples are presented. [ABSTRACT FROM AUTHOR]

In this paper, we present novel finite two-term cosine time functions, whose frequency responses exhibit triple pass bands, and these bands have the narrowest possible null-to-null lobe width ( ± 1 / T , where T is the duration of the time function), equal to that of a rectangular window. More importantly, these functions provide much higher decay of sidelobes (18 dB/octave) than that is obtainable (6 dB/octave) from the rectangular window. Therefore these functions can be used for designing triple band surface-acoustic-wave (SAW) filters, or for synthesizing the finite-impulse-response filters requiring triple bands. As an example, a triple band SAW filter centered at 25 MHz is designed employing two interdigital transducer structures. The modeling of the electrode-overlap pattern as a sampled data structure is explained, and the simulated frequency plot of the triple band SAW filter is shown. [ABSTRACT FROM AUTHOR]

In this paper, a novel pitch detection algorithm (PDA) is proposed. Actually, pitch detection is a classical problem that has been investigated since the very beginning of speech processing. However, the novelty of the proposed method consists in establishing an empirical relationship between fundamental frequency ( f 0 ) and instantaneous frequency ( f i ), which serves as a basis to develop the proposed PDA. Even though f 0 and f i are defined as attributes of two different transforms, i.e., the Fourier transform and the Hilbert transform, respectively, the relationship proposed in this paper shows some interaction between both of them, at least empirically. The first step of this work consists in validating the proposed relationship on a large set of speech signals. Then, it is leveraged to develop an algorithm capable to (a) detect voiced/unvoiced parts of speech and (b) extract f 0 contour from f i values in the voiced parts. For evaluation purposes, the yielding f 0 contour is compared to some well-rated state-of-the-art PDA's. The main findings show that the quality of pitch detection obtained by the proposed technique is as satisfactory as some of top PDA's, either in clean or in simulated noisy speech. In addition, one of the main advantages consists in bypassing the traditional short-time analysis required to assume local stationarity in speech signal. [ABSTRACT FROM AUTHOR]

Autonomous cars, cloud computing and image recognition technology have spread drastically in recent years. However, operation of these applications requires a large number of multipliers and has a high hardware cost. Logarithmic converters can replace the multipliers at the expense of a large conversion error. This paper proposes a design for area-delay-product (ADP)-efficient logarithmic converters that use compensation methods involving mantissa bits. Replacing the conventional constant adder, bit-level manipulation and shift-and-add scheme, the proposed ADP-efficient logarithmic converters use mantissa bits to compensate for the conversion error of the accurate logarithmic curve. According to different specifications and applications of tolerable logarithmic conversion error, three sets of logarithmic converters, that is, methods 1, 2 and 3, were proposed to approximate the accurate logarithmic curve. The proposed ADP-efficient logarithmic converters can achieve a smaller implementation area, shorter latency and lower error than other methods in the literature. The total conversion errors of the proposed ADP-efficient logarithmic converters are 0.0379, 0.0303 and 0.0241 for methods 1, 2 and 3, respectively. Simultaneously, the proposed ADP-efficient logarithmic converters can attain ADP savings of 35.32, 69.71 and 66.19% compared to previous papers, for the proposed methods 1, 2 and 3. The design's very-large-scale-integration (VLSI) hardware implementation is synthesized by the Taiwan Semiconductor Manufacture Company using 0.18 μm CMOS technology. Additionally, the proposed ADP-efficient implementation structure is ROM-free, fast, simple and has a lower logarithmic conversion error range. The proposed converters can be easily implemented by VLSI digital circuits and may be applied to 5G technology and digital signal processing. [ABSTRACT FROM AUTHOR]

This paper proposes an adversary-resilient communication-efficient distributed estimation algorithm for time-varying networks. It is a generalization of the doubly compressed diffusion least mean square algorithm that is not adversary-resilient. The major drawback in existing adversary detectors in the literature is that they suggested the detection criterion heuristically. In this paper, an adversary detector is suggested theoretically based on a Bayesian hypothesis test (BHT). It is proved that the test statistics of the detectors is a distance metric compared to a threshold similarly to related papers in the literature. Hence, we prove the validity of the detection criterion based on BHT. The other difficulty encountered in existing works is the determination of thresholds. In this paper, the optimum thresholds are derived in closed form. Since the optimum thresholds need the values of unknown parameters, it is not feasible to derive them. Hence, suboptimal procedures for determining the thresholds are provided. Moreover, the convergence of the mean of the algorithm is investigated analytically. In addition, the Cramer–Rao bound of the problem of distributed estimation based on all node observations in the presence of adversaries is calculated. The simulation results show the effectiveness of the proposed algorithms and demonstrate that the proposed algorithms reach the performance of the algorithm when the adversaries are ideally known in advance, with some delay. [ABSTRACT FROM AUTHOR]

This paper presents complete design guidelines for a typical fractional-order Colpitts oscillator (FOCO) with a non-ideal op-amp. These guidelines include effects of op-amp non-idealities like, open-loop dc gain, unity gain frequency and output resistance, into the design. The relation among these non-idealities and the frequency of oscillation ( f o ) are analytically established and an optimal configuration for FOCO is identified. The paper discusses multiple case studies on parametric variation, parameter sensitivities and oscillator stability. The proposed guidelines are validated with LTSpice simulations and practical experimentation along with step-by-step design examples. Practical results are compared with earlier reported FOCO which shows higher frequency and larger amplitude for the designed FOCO at a comparable THD. The proposed guidelines are generic in the sense that they are applicable for any order (including integer order), for any type of fractor realization technique, and also for any active element which has first-order behavioral model. [ABSTRACT FROM AUTHOR]

When we study the Karnaugh map in the switching theory course, we learn that the ones in the map are combined in rectangles whose length and width must be a power of two. The result is the logic function described as a sum of products. This paper shows that we can also make groups where the length and width of the rectangles are equal to three. This results in a logic function that is simpler than the sum of products in terms of logic gates, leading to more hardware-efficient circuits. This idea is extended later in the paper to other groups of elements. Finally, a new perspective on the Karnaugh map that integrates the proposed approach with the conventional one is provided. This can be used in switching theory courses to improve the explanation of the Karnaugh map. [ABSTRACT FROM AUTHOR]

The problem of global asymptotic stability (GAS) of fixed-point state-delayed uncertain discrete-time systems combined with a saturation operator in state-space is investigated in this paper. The uncertainties in the system are presumed to be norm-bounded, which has been frequently employed in robust control for uncertain systems. This paper proposes a new GAS criterion for the considered system. The saturation nonlinearities associated with the system, which operate exclusively in the linear region, are identified using a unique methodology. Such identification leads to an improved characterization of saturation nonlinearities involved in the system. The proposed criterion utilizes an upper bound of parameter uncertainties, an asymptotic bound on the system's states, and a precise characterization of saturation nonlinearities involving system parameters. The merit of the criterion is also exemplified through examples along with simulation results. [ABSTRACT FROM AUTHOR]

Convolutional neural network (CNN) has emerged as one of the primary tools for image classification. In particular, deep CNNs are now considered the main tool for this purpose. However, these networks are often large and require computing and storage power that may not be available in very small sensor devices such as IoT (Internet of Things) devices; their training is also time and computing power consuming. As a result, in some applications, reducing the size of input data (images) and the processing network becomes necessary. Such reduction usually comes at the cost of reduced classification performance. In this paper, we consider networks with under 200k learnable parameters, as opposed to millions in deeper networks. We examine how domain transforms can be used for efficient size reduction and improvement of classification accuracy for small networks. We emphasize that finding optimal hyperparameters or network configurations is not our objective in this paper. It is shown that by using transforms such as discrete wavelet transforms (DWT) or discrete cosine transform (DCT), it is possible to efficiently improve the performance of size-reduced networks and inputs. We demonstrate that in most cases, the improvement can be traced to higher entropy of resized input using transforms. While transforms such as DCT allow variable input and network sizes to be utilized, DWT proves to be very effective when significant size reduction is needed (improving the result by up to 5%). It is also shown that input size reduction of up to 75% is possible, without loss of classification accuracy in some cases. We use two datasets of small images, including Fashion MNIST and CIFAR-10, to evaluate the performance of size reduction methods. [ABSTRACT FROM AUTHOR]

In this work, an efficient full-swing (FS)-gate diffusion input (GDI) logic style is used for implementing full adder (FA) and arithmetic logic unit (ALU) circuits. The performance of the ALU in delay time, power consumption, and area terms is strongly dependent on the performance efficiency of the FA circuits utilized in its construction. In this research paper, multi-efficient power and high-speed ALU circuits are proposed with outputs in full-swing form and improved area. The proposed circuit is evaluated and tested using the Cadence Virtuoso simulation package in the 65 nm CMOS process. The proposed ALUs reduce power consumption by 20.01% and 4% compared with the ALU implemented by the traditional CMOS technique and GDI-based unit, respectively. Also, four ALU designs utilizing four efficient FA circuits are presented. The proposed ALUs provide more than 74.4–23.3% improvement in the power-delay product (PDP) term compared with existing circuits. The simulation experiments reveal that with the proposed ALUs, the circuits achieve superior performance compared with previous and existing ALU designs, and the power efficiency of these units improves in a range of 18.7–20.1% over the CMOS-based units. Also, the various technology nodes and their impact are considered in this paper. [ABSTRACT FROM AUTHOR]

While the practical application of adaptive filters has indeed garnered substantial attention, two pressing issues persist that have a profound impact on their performance—system sparsity and the presence of contaminated Gaussian impulsive noise. In this research paper, we propose a novel approach to tackle both of these issues simultaneously by introducing the concept of a proportionate matrix. Specifically, we present a proportionate maximum total complex correntropy algorithm based on the errors-in-variables model. The paper presents a theoretical analysis of the steady-state weight error power under the influence of impulsive noise. Furthermore, it discusses the performance comparison in system identification and highlights the robustness of the proposed algorithm. To validate its effectiveness, a simulation involving stereophonic acoustic echo cancellation is conducted, and the results confirm the clear advantages of the proposed Proportionate Maximum Total Complex Correntropy algorithm. [ABSTRACT FROM AUTHOR]

This paper proposes an algorithm to obtain the sampling factors to model any frequency partitioning so that it is realized using low complexity rational decimated non-uniform filter banks (RDNUFBs). The proposed algorithm is employed to approximate the Bark scale to find a rational frequency partitioning that can be realized using RDNUFBs with less approximation error. The proposed Bark frequency partitioning is found to reduce the average deviation in centre frequency and bandwidth by 65.04% and 48.50%, respectively, and root-mean-square bandwidth deviation by 54.46% when compared to those of the existing perceptual wavelet approximation of Bark scale. In the first filter bank design approach discussed in the paper, a near-perfect reconstruction partially cosine modulated filter bank is employed to obtain the improved Bark frequency partitioning. In the second design approach, the hardware complexity of the PCM-based RDNUFB is reduced by deriving the channels with different sampling factors from the same prototype filter by the method of channel merging. There is a considerable reduction of 40.83% in the number of multipliers when merging of partially cosine modulated channels is employed. It is also found that both the proposed approaches can reduce the delay by 95.17% when compared to the existing rational tree approximation methods and hence, are suitable for real-time applications. [ABSTRACT FROM AUTHOR]

This paper studies global exponential stability (GES) of discrete-time (D-T) BAM neural networks (BAMNNs) affected by impulses and time-varying delays. An impulse-free D-T BAMNN with time-varying delays is first developed via the properties of M-matrices, and then the relation between the solutions of the original and new BAMNNs is established. From which, sufficient conditions for GES of the original BAMNN are derived by investigating GES conditions of the new BAMNN based on the counter-evidence and the nature of the nonsingular M-matrix. The results of illustrative examples show that the obtained GES criteria are effective. By comparing with previous results, this paper has the following three merits: (a) the obtained GES criteria are to check the positivity of eigenvalues of a constant matrix, which is easy to realize; (b) this method can be used for other D-T system models affected by impulses and time-varying delays; and (c) the numerical results show that our method is less conservative than ones in other literatures. [ABSTRACT FROM AUTHOR]

This paper presents a detailed design of a broadband power amplifier (PA) utilizing 0.18-μm CMOS technology, operating in the frequency range of 2–5 GHz. This ultra-wideband (UWB) power amplifier uses a cascode topology with current reuse to enhance gain over the desired frequency band, utilizing an inductive source degenerate design to achieve the best linearity and output power and gain while maintaining a wide bandwidth. The inductive source degenerate adopts bondwire and microstrip line design. The paper includes extensive analysis, simulation results, and measurements to verify the performance of the amplifier and demonstrate its suitability for broadband power amplifier applications. The measured results for the proposed UWB power amplifier show excellent performance, with an impressive third-order input intercept point (IIP3) ranging from 15 to 18.7 dBm over the frequency range of 2–5 GHz. The output third-order intercept point (OIP3) ranges from 33 to 35 dBm. Additionally, the gain flatness is specified as 12.5 ± 1.5 dB. The reverse isolation is greater than or equal to − 40 dBm. The PA has a P1dB output power (OP1dB) of 10.2 dBm at 3 GHz, 9.2 dBm at 4 GHz, and 9.2 dBm at 5 GHz, respectively. The PA has a P1dB power-added efficiency of 30.5% at 3 GHz, 23% at 4 GHz, and 22% at 5 GHz with a 50 Ω load termination. The power dissipation is 33.3 mW at a supply voltage of 1.8 V. In addition, the PA has an excellent small group delay variation of ± 65 ps. The chip area is 1.01 × 1.08 mm2, including the pads. The proposed UWB PA demonstrates exceptional linearity, efficiency, and group delay in the 2–5 GHz frequency band. [ABSTRACT FROM AUTHOR]

The journal "Circuits, Systems & Signal Processing" has announced a call for nominations for three Best Paper Awards. The awards include the Armen H. Zemanian Best Paper Award for the best paper in the area of Circuits and Systems, the Sydney R. Parker Best Paper Award for the best paper in the area of Signal Processing, and the M.N.S. Swamy Best Paper Award for the best paper in the previous two years regardless of the area. Nominations should be submitted to the Editor-in-Chief, M.N.S. Swamy, by July 31, 2024, and should include the title of the paper, names of the authors, page numbers, the issue in which it was published, and a paragraph explaining the reasons for the nomination. [Extracted from the article]

This paper investigates the sliding mode control (SMC) problem of discrete-time singular Markovian jump systems with time-varying delay under an event-triggered strategy (ETS). In the system we study, there are uncertainties in the system matrix, especially in the difference matrix. A state augmentation strategy is proposed due to uncertainties in the difference matrix. A suitable sliding mode surface (SMS) function is built, and the corresponding sliding mode dynamics (SMD) is derived based on ETS. Some sufficient conditions are presented to guarantee that SMD is H∞\documentclass[12pt]{minimal}\usepackage{amsmath}\usepackage{wasysym}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{amsbsy}\usepackage{mathrsfs}\usepackage{upgreek}\setlength{\oddsidemargin}{-69pt}\begin{document}$$H_{\infty }$$\end{document}-admissible for all uncertainties. Furthermore, an SMC law under ETS is derived so that the system trajectories can be driven onto the prescribed SMS in finite time. Finally, the effectiveness of the approach in the paper is demonstrated by two simulated examples. [ABSTRACT FROM AUTHOR]

In this paper, we propose a parameters estimation method for three-dimensional (3-D) mixed far-field and near-field sources based on nonuniform cross tripole array where location and polarization estimates are considered. By designing several special fourth-order cumulants and making use of the shift-invariant properties in cumulant domain, the 2D-DOA and polarization information at five known tripoles are captured. With the 2D-DOA estimates, a matrix equation is derived to classify the source type and determine the coarse locate estimates. Then, the coarse parameters estimates can be used as references for disambiguating the cyclically ambiguous estimates to obtain the high accuracy estimation of 2D-DOA and range. Compared with the existing methods, The proposed one can separate and locate the mixed sources in 3-D space with the priori knowledge of only five sensors' positions. Moreover, it does not restrict the array sensors to be uniformly placed with inter-element spacing no more than a quarter or one-eighth wavelength, and thus, the array aperture can be extended by extending the inter-sensor spacing. Additionally, requirements for the placement of five known tripoles are analyzed. Numerical results show that the proposed algorithm achieves high parameter estimation accuracy. [ABSTRACT FROM AUTHOR]

Blind image deblurring aims to reconstruct a sharp version of a given blurred image by estimating the blur kernel. This is generally a highly ill-posed problem because both the blur kernel and the sharp image are unknown. To solve this problem, different methods use sophisticated image priors to construct an intermediate sharp image and then estimate the blur kernel. In this paper, we utilize the advantage of the pyramid scheme that is usually used in blind deblurring literature, and instead of using complicated constraints to reconstruct the intermediate latent image, we use a super-resolution approach in a scale-by-scale manner. This scheme is performed quickly, although we realized that multi-scale prior leads to noisy kernels, and the estimated blur kernels are not sufficiently continuous. We focus on kernel regularization to encourage sparsity and continuity, thus we add a kernel continuity constraint to the optimization function. Finally, by conducting experiments and providing quantitative and visual results, we demonstrate that the proposed deblurring method is fast and provides acceptable results for blurred cases where state-of-the-art methods fail. [ABSTRACT FROM AUTHOR]

A joint empirical mode decomposition and optimal frequency band estimation approach is proposed to reduce the low-frequency noise adaptively. In each level of the decomposition, this paper considers the residue as the approximation of the low-frequency noise while the sum of the intrinsic mode functions is considered as the approximation of the denoised signal. Conventionally, the signal-to-noise ratio (SNR) of the denoised signal is used as a criterion for determining the optimal level of the decomposition. In particular, the level of the decomposition which achieves the highest SNR is defined as the optimal level of the decomposition. However, as the clean signal is unknown, it is difficult to find the optimal level of decomposition. To address this issue, the cutoff frequency of the residue in each level of the decomposition is estimated via an optimization problem. Here, the total energy of the difference between the energy spectrum of the residue and that of its corresponding matched filter is minimized subject to a constraint on the energy of the matched filter being equal to that of the residue. After finding the solution of the optimization problem for each level of the decomposition, the index corresponding to the global minimum of the absolute difference of the optimal cutoff frequency between two consecutive levels of the decomposition is defined as the optimal level of the decomposition. Since the optimal level of the decomposition is found automatically, our proposed method avoids the inappropriate selection of the parameter. Besides, the computer numerical simulation results show the superior performance yielded by our proposed method over the existing approaches on various test signals at different noise levels. [ABSTRACT FROM AUTHOR]

The research on dissipative-chaotic systems has exhibited plenty of remarkable results for past few decades. But unfortunately, the attackers can quickly rebuild the attractors in dissipative systems, causing multiple information security issues. Compared to the dissipative system, the conservative system can efficiently prevent such external attacks due to the presence of flows without any attractors. Therefore, the study of conservative chaotic systems is essential for secure communications and engineering applications. Presently, a few studies are there for non-autonomous conservative systems based on simple memristor. For this reason, a circuit-based new simple memristor is designed and using this, a non-autonomous four-dimensional hyperchaotic conservative system without an equilibrium point is proposed in this paper. The hyperchaotic behavior and multiple offset boosting are explored by phase portraits, bifurcation diagrams, and Lyapunov exponent plots. It is found that due to the presence of applied stimuli, the proposed system also produces time-varying hyperchaos and offset boosting. Furthermore, different torus structures, triggered by initial conditions and control parameters are observed here as the prominent property of the conservative system. An analog circuit of the chaotic system is developed and executed experimentally to verify the correctness of the numerical model. Concurrently, a new algorithm for image encryption using the proposed hyperchaotic map is stated and the efficiency of this algorithm is confirmed by several security tests. [ABSTRACT FROM AUTHOR]

Speech emotion recognition (SER) is one of the most important approaches for emotional human–computer interaction. Hence, in recent years, SER has received a great deal of attention from many researchers and has become one of the most interesting and challenging topics. In most SER systems, the extraction of ineffective features is the most challenging problem. Thus, in this paper, a novel emotion recognition system was proposed based on the reconstructed phase space of speech. In this method, the three-dimensional reconstructed phase space of the speech signals was calculated. Consequently, emotion-related patterns formed in this space were converted into 3D tensors. Then, a three-dimensional convolutional neural network was employed to analyse the patterns and classify the corresponding emotions. Results of the experiments on the two public datasets including Berlin EMO-DB and eNTERFACE05 showed the promising functioning of the proposed scheme, which could significantly improve the performance of speech emotion recognition in contrast to other studies. [ABSTRACT FROM AUTHOR]

Speech is produced by exciting time-varying vocal tract with time-varying impulse-like excitations called epochs. In the literature, epoch extraction methods performed well on clean speech, but, detecting epoch locations from the band-limited signal like telephonic speech is difficult due to loss of information at low frequencies. This paper proposes a Stockwell transform (S-Transform)-based method that can find epochs accurately from the telephonic speech. The frequency-dependent Gaussian window and localization capabilities of S-Transform will reduce the effect of the bandpass nature of the telephonic channel. The telephonic channel is simulated using a 300–3400 Hz bandpass filter. The proposed method is evaluated on five speakers data, namely BDL, SLT, JMK, KED, and RAB, from CMU arctic database. The results are compared with the state-of-the-art methods for both clean speech and telephonic speech. The proposed method produced comparable results with existing methods on clean speech but has shown an improvement of 4.68 % over state-of-the-art methods. [ABSTRACT FROM AUTHOR]

In this paper, we investigate the h-stability analysis with respect to part of the variables of nonlinear non-autonomous systems. With the help of the notion of practical h-stable functions, input-to-state practical partial h-stability (h-ISppS), integral input-to-state practical partial h-stability (h-iISppS) and practical partial h-stability are considered. Moreover, some sufficient Lyapunov-like conditions are derived to check the partial input-to-state practical h-stability of two important classes of nonlinear systems, namely perturbed and cascaded systems. Furthermore, two numerical examples are given to illustrate the effectiveness and the superiority of the results. [ABSTRACT FROM AUTHOR]

The canonical polyadic decomposition (CPD) of higher-order tensors, a.k.a. PARAFAC, has shown excellent performance in two-dimensional direction of arrival (DOA) estimation using the acoustic vector sensor array (AVSA). However, most existing studies pay little attention to the manifold matrix structure of the AVSA during the CPD and are designed for uncorrelated sources. This paper presents a constrained CPD-based algorithm for DOA estimation using a uniform linear AVSA, whose manifold matrix is highly structured. Specifically, the manifold matrix equals to the Khatri-Rao product of a Vandermonde matrix and a proportional column-norm matrix. We show that DOA estimation accuracy is further improved by incorporating the prior structured information. Besides, we also extend the Toeplitz decorrelation technique to the AVSA to handle possibly correlated sources. The algorithm does not require iteration or peak searching and thus is computationally effective. Numerical simulations verify the effectiveness and superior performance of the algorithm. [ABSTRACT FROM AUTHOR]

In this paper, we introduce a novel statistical image watermark decoder based on robust nonsubsampled Shearlet transform (NSST)-polar linear Canonical transform (PLCT) magnitude and effective vector shifted exponential normal mixtures-based hidden Markov tree (vector SENM-HMT). We begin with a detailed study on the robustness and statistical characteristics of local NSST–PLCT magnitudes of natural images. This study reveals the excellent robustness, highly non-Gaussian marginal statistics and strong dependencies of local NSST–PLCT magnitudes. We also find that conditioned on their generalized neighborhoods, the local NSST–PLCT magnitudes can be approximately modeled as shifted exponential normal variables. Meanwhile, model parameters can be estimated effectively by using density-preserving hierarchical EM (DPHEM) and upward–downward approach. Based on these findings, we model local NSST–PLCT magnitudes using a vector SENM-HMT that can capture all interscale, interdirection and interlocation dependencies. Finally, we develop a new statistical image watermark decoder using the vector SENM-HMT and maximum likelihood (ML) decision rule. Experimental results demonstrate the efficacy and superiority of the proposed approach. [ABSTRACT FROM AUTHOR]