Defines four types of fuzzy neurons and proposes the structure a four-layer feedforward neural network (FNN) its associated learning algorithm. The proposed FNN performs well when used to recognize shifted distorted training patterns. When an input pattern is provided, first fuzzifies this then computes similarities all learned reaches conclusion by selecting with highest similarity gives nonfuzzy output. 26 English alphabets 10 Arabic numerals, each represented 16/spl times/16 pixels, were...

Traffic networks exhibit complex spatial-temporal dependencies, and accurately capturing such dependencies is critical to improving prediction accuracy. Recently, many deep learning models have been proposed for dependency modeling. While numerous developed modeling, most rely on different types of convolutions extract spatial temporal correlations separately. To address this limitation, we propose a novel framework traffic called GraphSAGE-based Dynamic Spatial-Temporal Graph Convolutional...

Common spatial pattern (CSP) is an efficient algorithm widely used in feature extraction of EEG-based motor imagery classification. Traditional CSP depends only on filtering, that aims to maximize or minimize the ratio variances filtered EEG signals different classes. Recent advances approaches show temporal filtering also preferable extract discriminative features. In view this perspective, a novel spatio-temporal strategy proposed paper. To improve computational efficiency and alleviate...

Color correction and enhancement for underwater images is challenging due to attenuation scattering. The often have low visibility suffer from color bias. This paper presents a novel method based on filter array (CFA) an Retinex with dense pixels adaptive linear histogram transformation degraded color-biased images. For any digital image in the RGB space, which captured by camera CFA, their values are dependent coupled because of interpolation process. So we try compensate red channel green...

DNA sequence analysis using digital signal processing requires conversion of a base to numerical sequence. The choice the representation affects how well its biological properties can be reflected in domain for detection and identification characteristics special regions interest. This paper presents some selected methods analysis, discusses their relative merits demerits, includes concluding remarks.

In this paper, a novel algorithm is proposed to design sparse FIR filters. It known that problem highly nonconvex due the existence of -norm filter coefficient vector in its objective function. To tackle difficulty, an iterative procedure developed search potential sparsity pattern, which then used compute final solution by solving convex optimization problem. each step, original successively transformed simpler subproblem. can be proved under weak condition, globally optimal solutions these...

A simple sigmoid-like second-order piecewise activation function suitble for direct digital hardware implementation is presented. Simulation results on the uses of and bipolar sigmoid training multilayer feedforward networks using backpropagation algorithm show that they have similar generalisation properties while has a slight advantage in convergence speed.

This paper presents a weighted least squares (WLS) method for IIR digital filter design using new stability constraint. Utilizing the reweighting technique, an iterative second-order cone programming (SOCP) is employed to solve problem, such that either linear or constraints can be further incorporated. In order guarantee of designed filters, constraint with prescribed pole radius derived from argument principle (AP) complex analysis. As compared other frequency-domain constraints, AP-based...

In this paper, two novel algorithms are developed to design sparse linear-phase (LP) FIR filters. Compared traditional methods, they can jointly optimize coefficient sparsity and order of an LP filter, so as achieve a balance between filtering performance implementation efficiency. The problem under consideration is formally cast regularized l <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">0</sub> -norm minimization problem, which then tackled...

In this paper, a novel method for IIR digital filter design using iterative second-order cone programming (SOCP) is proposed under the minimax criterion. The convex relaxation technique utilized to transform original nonconvex problem into an SOCP problem. By solving relaxed problem, lower and upper bounds on optimal value of can be obtained. order reduce discrepancy between problems, procedure developed. At each iteration, linear constraint further incorporated guarantee convergence...

In this paper, a novel method for IIR digital filter design using iterative second-order cone programming (SOCP) is proposed under the minimax criterion. The convex relaxation technique utilized to transform original nonconvex problem into an SOCP problem. By solving relaxed problem, lower and upper bounds on optimal value of can be obtained. order reduce discrepancy between problems, procedure developed. At each iteration, linear constraint further incorporated guarantee convergence...

This paper presents two-step design methodologies and performance analyses of finite-impulse response (FIR), allpass, infinite-impulse (IIR) variable fractional delay (VFD) digital filters. In the first step, a set (FD) filters are designed. second these FD filter coefficients approximated by polynomial functions FD. The FIR problem is formulated in peak-constrained weighted least-squares (PCWLS) sense solved projected (PLS) algorithm. For allpass IIR filters, nonconvex global solution...

A new proportionate affine projection sign algorithm is proposed for network echo cancellation. It uses a recursive procedure and takes into account the previously computed coefficients. shown that can obtain lower steady-state misalignment than other algorithms different paths, impulsive interferences step sizes.

A new algorithm for designing multilayer feedforward neural networks with single powers-of-two weights is presented. By applying this algorithm, the digital hardware implementation of such becomes easier as a result elimination multipliers. This proposed consists two stages. First, network trained by using standard backpropagation algorithm. Weights are then quantized to values, and slopes activation functions adjusted adaptively reduce sum squared output errors specified level. Simulation...

A multilayer feedforward neural network is applied to pulse compression. The 13-element Barker code and the maximum-length sequences (m-sequences) with lengths 15, 31, 63 b were used as signal codes, four networks implemented, respectively. In each of these networks, number input units was same length while hidden three output one. training backpropagation learning epochs 500. Using this approach, a more than 40 dB peak signal-to-sidelobe ratio can be achieved. These fault-tolerant provide...

In this paper, we propose a novel algorithm for sparse finite impulse response (FIR) filter designs. The objective of the digital design problem considered in paper is to reduce number nonzero-valued coefficients, subject weighted least-squares (WLS) approximation error constraint imposed on frequency domain. proposed method inspired by iterative shrinkage/thresholding (IST) algorithms, which are used and redundant representation signals. basic idea successively transform original nonconvex...

This paper presents a new algorithm using semidefinite programming (SDP) relaxation to design infinite impulse response digital filters in the minimax sense. Unlike traditional algorithms that try directly minimize error limit, proposed employs bisection searching procedure locate minimum limit of approximation error. Given fixed at each iteration, SDP technique is adopted formulate problem convex form. In practice, true cannot be always obtained. Thus, regularized feasibility procedure. The...

In this paper, four fuzzy filters for filtering images contaminated with random, impulse, and sum of random impulse noises are introduced. each these filters, the output pixel a filtered image at center moving window area is defined as normalized weighted input pixels within window. Simulation results indicate that some show improvement over standard median average in reducing three noises.

In order to make good use of the linearity, necessity, and sufficiency stability-triangle-based stability conditions, a sequential minimization procedure is proposed in this brief convert minimax design an infinite impulse response (IIR) filter sequence subproblems. Aside from numerator, only one second-order factor denominator optimized each The then combined with Levy-Sanathanan-Koerner-based direct method for Many IIR filters have been designed using procedure. Three examples are provided...

Based on the biorthogonal analysis approach, a multiwindow real-valued discrete Gabor transform (M-RDGT) for periodic sequences is presented to efficiently analyze dynamic time-frequency content of signal containing components with multiple and/or time-varying frequencies. The M-RDGT offers computationally efficient implementation as well formulation complex-valued (M-CDGT). completeness condition proved be equivalent its biorthogonality constraint between windows and synthesis windows. can...

In this paper, fuzzy inference models for pattern classifications have been developed and networks based on these are proposed. Most of the existing rule-based systems difficulties in deriving rules membership functions directly from training data. Rules obtained experts. Some approaches use backpropagation (BP) type learning algorithms to learn parameters However, BP take a long time converge they require an advanced setting number rules. The work determine demands lots experiences...

Digital processing of a nucleotide sequence requires it to be mapped numerical in which the choice numeric mapping affects how well its biological properties can preserved and reflected from domain domain. spectral analysis sequences unfolds period-3 power value is more prominent an exon as compared that intron sequence. The success based classification depends on threshold value. main purposes this article are introduce novel codes for 1-sequence representations compare them existing...