Search

Under the condition of low signal-to-noise ratio, for the problem of insufficient speech feature extraction and speech enhancement effect of the traditional neural network, this paper is based on empirical mode decomposition (EMD), temporal convolutional network (TCN), and gated convolution recurrent neural network (GCRN), while combining with feature fusion module (FFM), the adaptive mean median-empirical mode decomposition-multilayer gated feature fusion module convolutional recurrent neural networks (ME-MGFCRNs) for speech enhancement modeling. The network model uses a split-frequency learning strategy to learn low-frequency features and high-frequency features, i.e., the TCN and MGFCRN networks are used to obtain low-frequency and high-frequency features, and FFM processes the two sets of features to achieve speech enhancement in the form of feature mapping. The model proposed in this paper performs ablation and comparison experiments on the dataset to evaluate the enhancement effect of speech using PESQ, FwSegSNR, and STOI metrics. The research shows that under different noise environments and SNR conditions, the model proposed in this paper improves compared with other baseline models, especially under the low SNR condition of − 5 dB, FwSegSNR and PESQ improve by more than 0.86 dB and 0.02 compared with other baseline models. [ABSTRACT FROM AUTHOR]

The complex domain adaptive filtering algorithms has shown excellent performance in the field of signal processing. Among them, the well-known complex least mean square (CLMS) algorithm has been widely used in practical projects. However, the CLMS algorithm only considers the case where the output signal is disturbed by noise, ignores the input signal may also be interfered by noise. Moreover, the signal is likely to be disturbed by impulsive noise in practice. Based on the above problems, the complex total least mean M-estimation (CTLMM) adaptive algorithm based on the M-estimation function is put forward in this paper. The algorithm can well handle the case where both input and output signals are subject to complex domain noise interference, and it also shows robustness with impulsive noise. This paper also analyzes the local stability of the CTLMM algorithm and discuss the theoretical steady-state error of the algorithm. Finally, simulation results verify the advantages of CTLMM algorithm and the rightness of the theoretical analysis. [ABSTRACT FROM AUTHOR]

Constrained least mean square (CLMS) algorithm is the most popular constrained adaptive filtering algorithm due to its simple structure and easy implementation. However, its convergence slows down when the input signal is colored. To address this issue, this paper firstly introduces the normalized subband adaptive filter (NSAF) into the constrained filtering problem and derives a constrained NSAF (CNSAF) algorithm using the Lagrange multiplier method. Benefiting from the good decorrelation capability of the NSAF, the proposed CNSAF algorithm significantly improves the convergence performance of the CLMS algorithm under colored inputs. Then, the mean and mean-square stability of the CNSAF algorithm is analyzed, and the theoretical models to characterize the transient and steady-state mean square deviation (MSD) behaviors of the CNSAF algorithm are derived utilizing the Kronecker product property and vectorization method. Further to extend the CNSAF algorithm to the problem of sparse system identification, a sparse version of the CNSAF algorithm (S-CNSAF) is derived. Finally, the validity of the derived theoretical MSD prediction models and the superiority of the proposed algorithms are confirmed by extensive computer simulations on system identification with colored inputs. [ABSTRACT FROM AUTHOR]

The controlled autoregressive autoregressive moving average (CARARMA) models are of popularity to describe the evolution characteristics of dynamical systems. To overcome the identification obstacle resulting from colored noises, this paper studies the identification of the CARARMA models by forming an intermediate correlated noise model. In order to realize the real-time prediction function of the models, the on-line identification scheme is developed by constructing the dynamical objective functions based on the real-time sampled observations. Firstly, a rolling optimization cost function is built based on the observation at a single sampling instant to catch the modal information at a single time point and a generalized extended stochastic gradient (GESG) algorithm is proposed through the stochastic gradient optimization. Secondly, a rolling window cost function is built in accordance with the dynamical batch observations within data window by extending the proposed GESG algorithm and the multi-innovation generalized extended stochastic gradient algorithm is derived. Thirdly, from the perspective of theoretical analysis, the convergence proof of the proposed algorithm is provided based on the stochastic martingale convergence theory. Finally, the simulation analysis and comparison studies are provided to show the performance of the proposed algorithms. [ABSTRACT FROM AUTHOR]

In this paper, we present an underdetermined estimation algorithm for two-dimensional (2-D) direction of arrival (DOA) based on three-parallel co-prime arrays. By exploiting the correlation among the received signals in both the time and spatial domains, a conjugate augmented spatial–temporal virtual array is constructed, which significantly increases the degrees of freedom (DOF) and expands the array aperture, effectively filling the holes within the physical array. By vectorizing the covariance matrix obtained from the spatial–temporal virtual array, we can transform the 2-D DOA estimation problem into a one-dimensional (1-D) sparse recovery problem. Thereafter, the excellent gridless methods: sparse and parametric approach (SPA) and the matrix pencil and pairing (MAPP), are adopted to efficiently resolve the 1-D sparse recovery problem, enabling the automatic pairing of the 2-D angles. And simulation results are provided to validate the enhanced accuracy of the proposed algorithm in comparison with other existing methods. [ABSTRACT FROM AUTHOR]

This paper introduces a novel algorithm for optimizing the coefficients of the digital filters used in incremental delta-sigma analog-to-digital converters (IDSC). This algorithm is modified from constrained linear least squares (LS) to improve the signal-to-noise-and-distortion ratio (SNDR) of IDSC and minimize the oversampling rate (OSR) of the modulator, which enhances system speed and reduces power consumption. In the case of the same SNDR, the first-order IDSC with the proposed filter can reduce the OSR of the modulator by 50%, at least compared to the IDSC with the conventional filter. The second-order IDSC with the proposed filter can reach a higher SNDR of 106 dB with an OSR of 64. Considering the nonlinearity of the integrator, the SNDR of IDSC with the proposed filter is also 10 dB greater than that of the conventional filters and 3dB greater than that of the IDSC using the near-optimal algorithm filter. The experimental results indicate that the proposed filter possesses an excellent figure of merit of 0.028 pJ / conv . [ABSTRACT FROM AUTHOR]

This paper investigates the finite-time and fixed-time bipartite consensus problem of multiple Euler–Lagrange systems (MELSs) under directed interaction topologies, considering external disturbances. Two novel distributed hierarchical control algorithms are developed to address the challenging problems. Specifically, a new distributed estimator is designed to estimate the states of agents in finite/fixed-time. Moreover, based on the obtained estimations, the finite/fixed-time controller is presented to accomplish the bipartite consensus of the MELSs in the local control layer. By utilizing Lyapunov stability and the nearest neighbor-interactions theory, sufficient conditions for achieving the designed performance of the MELSs are derived. Simulation examples are provided to illustrate the effectiveness of the proposed algorithms. [ABSTRACT FROM AUTHOR]

The performance of a narrowband active noise control (NANC) system can be significantly degraded due to the frequency mismatch (FM). In this paper, the statistical performance of a typical FxLMS-based NANC system in the presence of FM is analyzed in detail. Difference equations governing the system dynamics and closed-form steady-state mean-square error expressions are derived and discussed. The stochastic analysis results reveal that the FM introduces small troublesome sinusoids into the NANC system. The controller has to track these sinusoids to minimize the residual noise, which leads to a serious performance deterioration. As a by-product, the optimal step sizes that minimize the effect of a relatively small FM are also derived. The findings significantly enrich our understanding of the stochastic behavior of the FxLMS algorithm in the presence of FM and also provide some useful information for NANC system design. Extensive simulations are conducted to confirm the validity of the analytical findings. [ABSTRACT FROM AUTHOR]

High-accuracy positioning plays an important role in future 5G-Advanced cellular networks. However, it is quite challenging to utilize 5G signals to estimate positioning parameters accurately under multipath environments. To establish a high-precision real-time location system (RTLS), which is suitable for indoor scenarios, advanced parameter estimation techniques with low computational complexity should be developed. This paper proposes a modified subspace-based time-of-arrival (TOA) estimation algorithm for 5G signals with ultra-high accuracy and reduced complexity under multipath environments. A novel smoothing scheme termed as grouped subcarrier smoothing (GSS) is proposed to reduce the dimension of the covariance matrix involved in typical subspace-based methods. Then, the computational complexity of the corresponding eigenspace calculation is reduced. The TOA estimation performance is enhanced by constructing multiple snapshots of the estimated channel frequency response vector in the frequency domain. Finally, the super-resolution TOA estimate can be obtained using only one orthogonal frequency division multiplexing (OFDM) symbol. Simulation results show that the proposed method achieves centimeter-level TOA estimation accuracy under multipath channels when the signal-to-noise ratio (SNR) is high. [ABSTRACT FROM AUTHOR]

Blind source separation (BSS), as a digital signal processing approach, focuses on estimating the underlying source signals from their linear mixtures without any prior information about the source signals and mixing matrix. Conventional methods for the BSS, however, are incapable of separating the complex-valued source signals. By leveraging the negative conjugate gradient to minimize the least mean square error reconstruction (LMSER) principle in complex domain, this paper proposes a collection of least-squares algorithms for complex-valued BSS (CBSS), including least-mean square (LMS)-type algorithms and recursive least-squares (RLS)-type algorithms. We demonstrate the availability of the proposed algorithms in both circular and non-circular source signals separation. Especially, the RLS algorithm for the CBSS without prewhitening is superior in cross-talking criterion to the others, as verified by computer simulations on artificial source signals. [ABSTRACT FROM AUTHOR]

In this paper, a stochastic model is presented for the nonparametric variable step-size normalized least-mean-square (NP-VSS-NLMS) algorithm. This algorithm has demonstrated potential in practical applications and hence a deeper understanding of its behavior becomes crucial. In this context, model expressions are obtained for characterizing the algorithm behavior in the transient phase as well as in the steady state, considering a system identification problem and Gaussian input data. Such expressions reveal interesting algorithm characteristics that are useful for establishing design guidelines and for the advancement of more refined algorithms. Simulation results for various operating scenarios ratified both the model's accuracy and the algorithm's superior performance relative to other recent and relevant algorithms from the literature. [ABSTRACT FROM AUTHOR]

A novel multi-combined-step-size normalized subband adaptive filtering algorithm is proposed. Different from the traditional combined-step-size method, by employing a variable mixing parameter on each subband, our designed algorithm is able to combine a large step size with a small one more effectively for each subband. The subband mixing parameters are derived from the variance of the noise-free a priori subband error signals. In this design, a noniterative shrinkage strategy is also utilized to estimate the noise-free a priori subband error signals. Moreover, the mean-square and steady-state performances of the proposed algorithm are studied. Simulation results illustrate that the proposed algorithm outperforms other algorithms mentioned in this paper in terms of tracking capability, convergence speed and steady-state error. [ABSTRACT FROM AUTHOR]

To enhance the performance of the diffusion robust adaptive filtering algorithm, this paper presents a novel loss function framework named the "Tanh cost function framework." By utilizing this framework, we are able to incorporate standard cost functions, devise novel cost functions, and introduce the associated diffusion algorithm. Moreover, the convergence of the algorithm is thoroughly analyzed. Furthermore, the proposed diffusion Tanh algorithm family is simulated in a system identification model to assess its performance, demonstrating superior performance compared to the standard diffusive algorithm. Notably, the DTHLMS within the diffusion tanh algorithm family exhibits superior performance over the mainstream robust diffusion algorithms. [ABSTRACT FROM AUTHOR]

In this paper, we investigate the signal reconstruction of M-channel oversampled graph filter banks (MOSGFBs) on sparse graphs. The analysis filter bank of MOSGFBs can be in polynomial or node-variant (NV) form. Given the analysis filter bank and sampling matrix, the synthesis procedure can be formulated into a quadratic optimization problem. To solve this problem, a novel distributed conjugate gradient (DCG) algorithm is proposed, in which the step size and increment involved in the conjugate gradient iteration are evaluated in a distributed manner. The local step sizes and increments are first calculated using data from neighboring nodes, and then patched to participate in the iteration. The DCG algorithm involves the exchange of information between neighboring nodes and executes the synthesis computation locally; therefore, the synthesis procedure has a low implementation complexity and is suitable for processing large-scale datasets. Numerical examples conducted on synthetic and real-world datasets demonstrate the effectiveness of the proposed algorithm. [ABSTRACT FROM AUTHOR]

Bright-dark components and edge details are the most important complementary information between infrared and visible images. To extract and fuse them efficiently, a novel non-subsampled morphological fusion algorithm is proposed in this paper. The algorithm uses non-subsampled pyramid (NSP) as the spatial-frequency splitter to decompose the source image to get a series of high-frequency detail images and one low-frequency background image. Then, a dual-channel multi-scale top–bottom hat (MTBH) decomposition is constructed to extract the bright-dark details from the low-frequency background. In addition, to extract the edge details with different directions from high-frequency images, a dual-channel multidirectional inner-outer edge (MIOE) decomposition is constructed. Through these decompositions, the bright-dark information and edge details present in the source images can be effectively extracted. Then, based on the distinct roles of the extracted information, the decomposed images are fused using diverse fusion strategies. Subsequently, the fused image is reconstructed using the appropriate inverse transforms corresponding to each decomposition. The experimental results demonstrate that the fusion images generated by this algorithm exhibit richer details and higher image contrast compared to those produced by state-of-the-art algorithms. [ABSTRACT FROM AUTHOR]

The time–frequency analysis (TFA) method is an effective tool to separate and extract main components for non-stationary signals such as vibration signals and seismic signals, which are time-varying and affected by high noise. However, suffering from the Heisenberg uncertainty principle and cross terms of time–frequency result, conventional TFA methods usually produce vague time–frequency representations (TFRs). As a branch of the TFA method, current redistributive compressive transforms enable to generate clear TFR. However, these techniques are limited to a singular type of signal, which is not applicable to deal with complicated signals in production. In order to enhance the applicability and the time–frequency (TF) aggregation capability, this paper proposes a promoted TFA method 2D-FTSST2 based on the synchrosqueezing transform combining two-dimensional information of time and frequency domains. For an accurate IF estimate, we also define a time redistribution operator, which can describe strong time and frequency-varying signals. This algorithm not only provides a high-resolution decomposition of multicomponent signals but also enables to extract main features in noisy environments. Experiments on simulated signals and real data confirm the validity and effectiveness of the proposed algorithm. [ABSTRACT FROM AUTHOR]

One of the main challenges in using adaptive filtering algorithms is efficiently emulating a system subject to noisy disturbances. This can be facilitated in applications where the system response to impulse is sparse, which allows for acceleration of the convergence rate if appropriate strategies are used. As a result, methods that impose norm constraints on the estimates are widely used. However, in the case of non-stationary plants to be identified, there is a gap in terms of theoretical performance guarantees of these algorithms. This paper proposes a novel stochastic model capable of predicting the performance of the ℓ0\documentclass[12pt]{minimal}\usepackage{amsmath}\usepackage{wasysym}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{amsbsy}\usepackage{mathrsfs}\usepackage{upgreek}\setlength{\oddsidemargin}{-69pt}\begin{document}$$\ell _0$$\end{document}-LMS algorithm in identifying a plant subjected to a first-order Markovian disturbance. Therefore, a tracking analysis is carried out, including both the average performance of the adaptive coefficients and second-order statistics of these coefficients. The theoretical model offers an analytical equation that predicts the asymptotic mean squared deviation in terms of the variance of the Markovian disturbance. Further, for most simulated scenarios, the theoretical model’s error in mean squared deviation remains below 1 dB, even when the learning step varies across several orders of magnitude. It was possible to observe that the theoretical model can accurately predict the steady-state regime for a wide range of learning step values and calculate an optimal value for this parameter. The findings are confirmed through extensive simulations. [ABSTRACT FROM AUTHOR]

Accurate measurement is an important prerequisite for sound source localization. In the enclosed environments, noise and reverberation tend to cause localization errors. To address these issues, this paper proposes a compound model particle filter algorithm based on multi-feature. Based on a multi-feature observation, the likelihood function of speaker tracking is constructed for particle filter, and multi-hypothesis and frequency selection function are adopted to establish multi-feature optimization mechanism, including time delay selection and beam output energy fusion. It is found that they effectively solved the difficulty in the simultaneous suppression of noise and reverberation by single feature. Moreover, considering the randomness of speaker motion, a compound model for sound source tracking is developed, where the stability of the speaker tracking system is improved by integrating multi-feature observation into the compound model filtering. The experimental results with both simulated and real acoustic data indicate that the proposed method has better tracking performance, compared with the existing ones with low SNR and strong reverberation as well as highly mobile conditions. [ABSTRACT FROM AUTHOR]

Aiming at the weak robustness of possibilistic fuzzy clustering against noise, a robust possibilistic fuzzy additive partition clustering with master–slave neighborhood information constraints is proposed for high noise image segmentation. This algorithm first constructs a master–slave neighborhood model, which consists of the master neighborhood window of the current pixel and the slave neighborhood window around the master neighborhood pixel. Then, the master–slave neighborhood information is integrated into the possibilistic fuzzy additive partition clustering model, and a novel robust possibilistic fuzzy clustering model incorporating deep local information is constructed. Next, this clustering model is further simplified by Cauchy inequality and a robust master–slave neighborhood information-driven possibilistic fuzzy clustering algorithm is derived by optimization theory. Extensive experimental results indicate that the proposed algorithm is very effective for noisy image segmentation, and its segmentation performance is significantly better than many existing state-of-the-art fuzzy clustering-related algorithms. In short, the work of this paper has profound significance for the development of robust fuzzy clustering theory. [ABSTRACT FROM AUTHOR]

In this research paper, we introduce an adaptive block-matching motion estimation algorithm to improve the accuracy and efficiency of motion estimation (ME). First, we present a block generation system that creates blocks of varying sizes based on the detected motion location. Second, we incorporate predictive tools such as early termination and variable window size to optimize our block-matching algorithm. Furthermore, we propose two distinct search patterns to achieve maximum quality and efficiency. We evaluated the proposed algorithms on 20 videos and compared the results with known algorithms, including the full search algorithm (FSA), which is a benchmark for ME accuracy. Our proposed quality-based algorithm shows an improvement of 0.27 dB in peak signal-to-noise ratio (PSNR) on average for reconstructed frames compared to FSA, along with a reduction of 71.66% in searched blocks. Similarly, our proposed efficiency-based method results in a 0.07 dB increase in PSNR and a 97.93% reduction in searched blocks compared to FSA. These findings suggest that our proposed method has the potential to improve the performance of ME in video coding. [ABSTRACT FROM AUTHOR]