The aim of multi-speaker emotional speech synthesis is to generate speech for a designated speaker in a desired emotional state. The task is challenging due to the presence of speech variations, such as noise, content, and timbre, which can obstruct emotion extraction and transfer. This paper proposes a new approach to performing multi-speaker emotional speech synthesis. The proposed method, which is based on a seq2seq synthesizer, integrates emotion embedding as a conditioned variable to convey exact emotional information from reference audio to the synthesized speech. To boost emotion representation capability, we utilize a three-dimensional acoustic feature as input. And an emotion generalization module with adaptive instance normalization (AdaIN) is proposed to obtain emotion embedding with high generalization ability, which also results in improved controllability. The derived emotion embedding from the generalization module can be readily conditioned by affine parameters, allowing for control both the emotion category and the emotion intensity of synthesized speech. Various emotional speech synthesis experimental results of the propposed method demonstrate its state-of-the-art performance in multi-speaker emotional speech synthesis, coupled with its advantage of high emotion controllability.

It is widely acknowledged that distinguishing genuine speech from spoofed speech encompasses various subbands and temporal segments within speech signals. However, prevailing spoofing detection methods tend to oversimplify the relationships between these cues by employing linear models. In this paper, we introduce a multi-level information aggregation Graph Attention Networks (MiaGATs) to generate highly discriminative features for fake speech detection (FSD). In MiaGATs, each subband and temporal segment of a speech signal is represented as distinct nodes. MiaGATs incorporates channel information aggregation within each node to effectively harness the unique spectral and temporal characteristics during the feature encoding stage. In particular, MiaGATs address the interactions between nodes through indirect node aggregation and integrates both indirect and direct node aggregation by max-pooling operation. Experimental results on ASVspoof2019 and ASVspoof2021 LA databases show significant relative improvement compared to the current state-of-the-art. In comparison to the leading integrated spectro-temporal graph attention networks, MiaGATs gains an impressive performance improvement in various conditions, underscoring MiaGATs's position as a new benchmark in spoofing detection performance.

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 spatial-temporal dependency modeling. While numerous deep learning models have been developed for spatial-temporal dependency modeling, most rely on different types of convolutions to extract spatial and temporal correlations separately. To address this limitation, we propose a novel deep learning framework for traffic prediction called GraphSAGE-based Dynamic Spatial-Temporal Graph Convolutional Network (DST-GraphSAGE), which can capture dynamic spatial and temporal dependencies simultaneously. Our model utilizes a spatial-temporal GraphSAGE module to extract localized spatial-temporal correlations from past observations of a node’s spatial neighbors. Meanwhile, the attention mechanism is incorporated to dynamically learn weights between traffic nodes based on graph features. Additionally, to capture long-term trends in traffic data, we employ dilated causal convolution as the temporal convolution layer. A series of numerical experiments are conducted on five real-world datasets, which demonstrates the effectiveness of our model for spatial-temporal dependency modeling.

Common spatial pattern (CSP) is widely used in motor imagery classification tasks. Classical CSP depends only on spatial filters. To improve its performance, a novel and efficient spatio-temporal filtering strategy is proposed in this paper to extract discriminative features. Common temporal filters are shared among all the spatial channels, so as to reduce the overfitting risk in the case of a small sample size. An efficient alternating optimization algorithm is also developed to optimize coefficients of spatial and temporal filters. To alleviate adverse effects of noise and artifacts and improve implementation efficiency, an ℓ1-norm-based sparsity regularization term is further introduced. The resulting problem is tackled by the reweighting technique. The effectiveness of the proposed algorithm is validated by the experiments using open datasets of BCI Competition.

In this paper, we develop a novel deep-network architecture for semantic segmentation. In contrast to previous work that widely uses dilated convolutions, we employ the original ResNet as the backbone, and a multi-scale feature fusion module (MFFM) is introduced to extract long-range contextual information and upsample feature maps. Then, a graph reasoning module (GRM) based on graph-convolutional network (GCN) is developed to aggregate semantic information. Our graph reasoning network (GRNet) extracts global contexts of input features by modeling graph reasoning in a single framework. Experimental results demonstrate that our approach provides substantial benefits over a strong baseline and achieves superior segmentation performance on two benchmark datasets.

In this paper, an adaptive modification rate artificial bee colony (AMR‐ABC) algorithm is proposed by incorporating a novel adaptive modification rate to adaptively balance exploration and exploitation to determine which parameters (or the number of parameters) to be updated in a solution during each iteration. The performance of the AMR‐ABC algorithm is compared to those the standard ABC algorithm and its two variants, and the Parks–McClellan algorithm for designing Type 3 (orders: 14, 26, and 38) and Type 4 (orders: 13, 25, and 37) linear phase FIR differentiators to evaluate their design capabilities. Design results have shown that the proposed AMR‐ABC algorithm (i) outperforms four other design algorithms with the lowest p ‐norm error in each of the six differentiator designs and (ii) is robust such that the same p ‐norm error solution with an equiripple amplitude response in each of the six differentiator designs can be obtained by repeating a design with a different population of randomly generated initial solutions. The filter coefficients of six linear phase FIR differentiator designs are given as benchmarks to compare the p ‐norm error performance of the AMR‐ABC algorithm to other algorithms. The AMR‐ABC algorithm is attractive to be used for optimisation in this and other design problems.

The multiwindow discrete Gabor transform (M‐DGT) is an effective time‐frequency analysis tool to analyse time‐varying signals containing components with multiple frequencies. In this study, fast block time‐recursive methods for computing the M‐DGT coefficients of a signal and the reconstruction of the signal from the transform coefficients are presented with steps as listed, respectively, in Algorithms 1 and 2, and their implementations using unified parallel lattice structures are also given. The proposed algorithms consisting of Algorithms 1 and 2 for respective forward and inverse transforms are compared to (i) those of the existing serial algorithms in terms of computational complexity and time; and (ii) those of the existing parallel algorithms in terms of hardware complexity. The results indicate that the proposed algorithm is fast in computing M‐DGT coefficients of a signal and reconstructing the signal with a reduced hardware complexity.

Graph filters are a fundamental tool in the field of graph signal processing. This paper focuses on the design of autoregressive moving average (ARMA) graph filters. In the proposed algorithm, the denominator part of an ARMA graph filter is decomposed as a cascade of a few second-order factors (SOFs) and a higher order factor (HOF), whose coefficients are updated sequentially in each iteration. In the proposed design algorithm, stability constraints are only imposed on the roots of SOFs and coefficients of the HOF are left unconstrained to enhance the design accuracy. Moreover, the number of SOFs can be automatically determined, which is convenient in practical applications. Simulation results demonstrate that the proposed algorithm can achieve higher computational efficiency and also approximation accuracy, compared to the state-of-the-arts of graph filters.

In this paper, a new algorithm is proposed for the design of sparse FIR filters. Traditional l1-optimization-based methods take all the coefficients into l1-norm minimization. However, it is unnecessary since some of them can only take nonzero values to satisfy design specifications. Furthermore, minimizing l1norm of all the coefficients could drive the design results to deviate from the optimal ones. The proposed algorithm aims to identify nonzero coefficients at some crucial positions in each iteration to minimize the number of nonzero coefficients. Simulation results demonstrate that the proposed algorithm can achieve better design results than traditional l1-optimization methods.

In this paper, a novel algorithm is developed for the minimax design of IIR digital filters. Using a partial second-order factorization (PSOF), the denominator polynomial of an IIR digital filter is decomposed as a cascade of second-order factors (SOFs) and a single higher-order factor (HOF). This is inspired by the fact that, when some poles are closer to the boundary of the stability domain, the other poles tend to stay inside the stability domain such that the specified frequency response can be best approximated. By means of the PSOF, stability constraints are only imposed on a limited number of SOFs and, thus, a better design could be attained. The proposed algorithm successively updates SOFs and HOF. To further reduce its computational complexity, the iterative weighted least-squares approach is applied to optimize each SOF or HOF. Simulation results demonstrate that the proposed algorithm can attain the balance between computational efficiency and design accuracy.

A novel stability condition is developed in this paper. It is both necessary and sufficient, which ensures that optimal design cannot be excluded from the admissible solutions. Compared to other necessary and sufficient stability conditions, the proposed one can be expressed as a quadratic constraint in terms of denominator coefficients, which facilitates its combination with other widely used IIR filter design strategies. In this paper, we adopt the Steiglitz-McBride scheme to design IIR filters. In each iteration, an approximation version of the proposed stability condition is further expressed as a set of linear inequality constraints, such that the resulting design problem becomes a quadratic program that can be efficiently and reliably solved. Simulations demonstrate the effectiveness of the proposed stability condition.

The length of windows used in the existing multiwindow complex-valued discrete Gabor transform (M-CDGT) is restricted to be the length of analyzed sequences. Consequently, if the lengths of analyzed sequences are long, to solve the windows requires high computation burden and memory and even leads to numerical instability. To overcome this problem, a multi-window real-valued discrete Gabor transform (M-RDGT) for long and infinite sequences is presented in this paper based on the biorthogonal analysis approach, and its corresponding biorthogonality constraint between analysis windows and synthesis windows is derived. The lengths of the analysis and synthesis windows are independent of the lengths of analyzed sequences so that one can apply finite (or short) windows to process any long (or even infinite) sequences. The completeness condition of the M-RDGT is proved to be equivalent to its biorthogonality constraint between analysis windows and synthesis windows. In addition, the M-RDGT can utilize the fast discrete Hartley transform algorithms for fast computation and has a simple relationship with the M-CDGT such that the M-CDGT coefficients can be directly computed from the M-RDGT coefficients.

Parallel‐computing‐based implementation of the two recent fast parallel algorithms for the discrete Gabor transform (DGT) is presented in this paper. First of all, the first existing block time‐recursive DGT algorithm with parallel lattice structure is analysed, and then an improved implementation method under a parallel computing environment is presented. Each parallel channel (i.e. process in parallel computing) in the improved method is independent, thereby reducing the interprocess communication by 99.2% on average over the original algorithm. Second, the second existing fast parallel DGT algorithm based on multirate filtering is analysed. Through the use of parallel computing, the communication overhead of the multirate filtering‐based parallel DGT algorithm is optimised and its time efficiency is raised from 31.26 times to 54.52 times faster than the serial fast DGT algorithm in processing of long sequences. Finally, the experimental results are compared and analysed, which indicate that the proposed fast DGT implementation methods are attractive for real‐time signal processing.

A large number of experiments have demonstrated that for an FIR filter the sparsity of filter coefficients is highly related to its filter order. However, traditional sparse FIR filter design methods focus on how to increase the number of zero-valued coefficients, but overlook the impact of filter orders on design performance. As an attempt to jointly optimize filter length and sparsity of an FIR filter, a novel method is proposed in this paper to design sparse linear-phase FIR filters. With peak error constraints, the objective function of the design problem is formulated as a combination of the sparsity of filter coefficients and a measure of the effective filter order. Then, the design problem is then recast as a weighted l0-norm optimization problem, which is solved by an efficient numerical method based on the iterative-reweighted-least-squares (IRLS) algorithms. Experimental results illustrate that the proposed method can efficiently reduce the effective filter order while enhancing the sparsity of an FIR filter.

In this paper, two proportionate affine projection sign algorithms are introduced. The performance of the proposed algorithms is compared with that of other proportionate affine algorithms under impulsive interference environment of a network echo cancellation system. It is shown that one of the proposed algorithms, termed memory improved proportionate affine projection sign algorithm (MIP-APSA), is the most robust to impulsive interferences and colored inputs. It is proved that MIP-APSA is a good candidate for network echo cancellation, because of its low complexity, good convergence speed and tracking abilities for echo paths with different sparseness measures, and projection orders.

In this paper, we present a novel algorithm to design sparse FIR digital filters in the minimax sense. To tackle the nonconvexity of the design problem, an efficient iterative procedure is developed to find a potential sparsity pattern. In each iteration, a subproblem in a simpler form is constructed. Instead of directly resolving these nonconvex subproblems, we resort to their respective dual problems. It can be proved that under a weak condition, globally optimal solutions of these subproblems can be attained by solving their dual problems. In this case, the overall iterative procedure can converge to a locally optimal solution of the original design problem. The real minimax design can then be achieved by refining the FIR filter obtained by the iterative procedure. The design procedure described above can be repeated for several times to further improve the sparsity of design results. The output of the previous stage can be used as the initial point of the subsequent design. Simulation results demonstrate the effectiveness of our proposed algorithm.

A novel algorithm is presented in this paper to design sparse FIR filters in the weighted least-squares (WLS) sense. The original design problem is cast as a constrained l0-norm optimization problem. To tackle the nonconvexity, an efficient iterative procedure is developed. In each iterative step, a subproblem in a simpler form is constructed. It can be demonstrated that in each iteration an optimal solution to each subproblem can be efficiently and reliably attained by the successive activation algorithm proposed in this paper, such that the overall design algorithm can converge to a local solution of the original design problem. Since its major part only involves scalar operations, compared with other sparse filter design approaches, the proposed design algorithm is computationally efficient. The effectiveness of the proposed design algorithm is demonstrated by numerical examples.

In the paper, the performances on FFT-based, FIR and IIR digital filter-based (DF-based) period-3 classifications of short exon and intron sequences of the human and eleven model organisms are studied. For each organism, sequences of 150-base length are chosen and 16,000 sequences are used for training and testing. FIR and IIR digital filter orders are chosen to be around a similar level of computational complexity as that of the FFT-based method. Also, three thresholds are considered. Simulation results indicate that the classification precisions obtained for different organisms are not the same probably due to their inherent genomic differences. Also, the FFT-based method, the FIR DF-based method (with filter order 24), and the IIR DF-based method (with filter order 6) can respectively yield top classifications in 9, 1, and 2 out of 12 organisms.

Novel algorithms for the multirate and fast parallel implementation of the 2-D discrete Hartley transform (DHT)-based real-valued discrete Gabor transform (RDGT) and its inverse transform are presented in this paper. A 2-D multirate-based analysis convolver bank is designed for the 2-D RDGT, and a 2-D multirate-based synthesis convolver bank is designed for the 2-D inverse RDGT. The parallel channels in each of the two convolver banks have a unified structure and can apply the 2-D fast DHT algorithm to speed up their computations. The computational complexity of each parallel channel is low and is independent of the Gabor oversampling rate. All the 2-D RDGT coefficients of an image are computed in parallel during the analysis process and can be reconstructed in parallel during the synthesis process. The computational complexity and time of the proposed parallel algorithms are analyzed and compared with those of the existing fastest algorithms for 2-D discrete Gabor transforms. The results indicate that the proposed algorithms are the fastest, which make them attractive for real-time image processing.

Novel and fast parallel algorithms for the 2-D DHT-based real-valued discrete Gabor transform (RDGT) and its inverse transform are presented based on 2-D filterbanks. A 2-D analysis filterbank is designed for the 2-D RDGT and a 2-D synthesis filterbank is designed for the 2-D inverse RDGT. The parallel channel in each of the two filterbanks has a unified structure and can apply the 2-D fast DHT (discrete Hartley transform) to reduce its computational load. The computational complexity of each parallel channel is very low and is independent of the oversampling rate. The computational complexity of the proposed parallel algorithms is analyzed and compared with that of the existing fastest parallel algorithms for the 2-D RDGT and its inverse transform. The results indicate that the proposed parallel algorithms are attractive for real tune image processing.

This research presents three new numerical representations for classifying short exon and intron sequences using discrete Fourier transform period-3 value. Based on the human genome, results indicate that the Complex Twin-Pair representation is attractive compared with other numerical representations and the approach has potential applications in genome annotation and read mapping.

Novel and fast parallel algorithms for the DHT-based real-valued discrete Gabor expansion (DGE) and discrete Gabor transform (DGT) are presented based on filterbanks. An analysis interbank is designed for the DHT-based real-valued DGT and a synthesis interbank is designed for the DHT-based real-valued DGE. The parallel channel in each of the two filterbanks has a unified structure and can apply the fast DHT (discrete Hartley transform) to reduce its computational load. The computational complexity of each parallel channel is very low and is independent of the oversampling rate. The computational complexity of the proposed parallel algorithms is analyzed and compared with that of the existing major parallel algorithms for the DGT and DGE. The results indicate that the proposed parallel algorithms are attractive for real time signal processing.

In this paper, a new iterative algorithm is proposed to design IIR digital filters in the minimax sense. Instead of directly minimizing the error limit of the approximation error, the proposed algorithm employs a bisection searching procedure to locate the minimum error limit. At each iteration, a feasibility problem with a given error limit is to be solved, which is constructed by applying the semidefinite programming (SDP) relaxation technique to transform the nonconvex approximation error into a convex form. In practice, however, the truly minimax solution cannot be always obtained by using this iterative procedure. Therefore, a regularization term needs to be incorporated in the objective of the feasibility problem at each iteration. Another bisection searching procedure is then deployed to find the minimum weight utilized in the regularized objective function of the feasibility problem. The stability of designed filters can be guaranteed by a monitoring strategy, which does not need to incorporate any other constraint to the formulation of the feasibility problem. The convergence of the proposed method can be guaranteed. The performances have been demonstrated by filter examples.

A two-stage design method of low-order fixed denominator IIR variable fractional delay (VFD) digital filters is presented in this paper. In the first stage, a set of FIR fractional delay (FD) filters are designed first. Each FIR FD filter design problem is formulated in the peak-constrained weighted least-squares (PCWLS) sense and solved by the projected least-squares (PLS) algorithm. Then, model reduction technique is applied on a time-domain average FIR filter to obtain the fixed denominator. The remaining numerators of the IIR FD filters can be obtained by solving linear equations derived from the orthogonality principle. In the second stage of the design, these FD filter coefficients are to be approximated by polynomial functions of FD. Three sets of filter-examples are given to illustrate the effectiveness of the proposed design method.

Two-dimensional fast Gabor transform algorithms are useful for real-time applications due to the high computational complexity of the traditional 2-D complex-valued discrete Gabor transform (CDGT). This paper presents two block time-recursive algorithms for 2-D DHT-based real-valued discrete Gabor transform (RDGT) and its inverse transform and develops a fast parallel approach for the implementation of the two algorithms. The computational complexity of the proposed parallel approach is analyzed and compared with that of the existing 2-D CDGT algorithms. The results indicate that the proposed parallel approach is attractive for real time image processing.

An iterative second-order cone programming (SOCP) approach is proposed in this paper. The original nonconvex design problem is first relaxed into an SOCP problem, which can provide a lower bound on the optimal value of the original problem. For reducing the gap between the original and the convex problem, an iterative procedure is developed. The initial point of the iterative procedure can be chosen as the solution obtained from the relaxed SOCP problem. Unlike other iterative approaches, the convergence of the proposed iterative procedure is definitely guaranteed. Design examples demonstrate the effectiveness of the proposed method.

The traditional DCT-based real-valued discrete Gabor transform (RGDT) was limited to the critical sampling case. The biorthogonality relationship between the analysis window and the synthesis window for the transform has not been unveiled. To overcome those drawbacks, this paper proposes a novel DCT-based real-valued discrete Gabor transform, which can be applied under both the critical sampling condition and the over-sampling condition. And the biorthogonality relationship between the analysis window and the synthesis window for the transform is also proved in this paper. Because it only involves real operations and can utilize fast DCT and IDCT algorithms for fast computation, it facilitates computation and implementation by hardware and/or software compared to the traditional complex-valued discrete Gabor transform.

In this paper, a novel state-space approach for model reduction of FIR digital filters is proposed. The method is based on the initial IIR filter obtained by the singular perturbational model reduction, and then utilizes the iterative algorithm via orthogonal projection under a different stop criterion to search a better result in L2norm sense. A lowpass filter example is designed to compare the performance of the proposed method and the singular perturbational model reduction in terms of frequency response and approximation error under the same reduced order. Results indicate that the proposed technique achieves comparable performance with much lower L2norm error than the singular perturbational model reduction.

This paper presents a novel adaptive wavelet denoising system for speech enhancement. The procedures of the standard wavelet thresholding are improved by optimizing the adaptive algorithms. The combination of the quantile based noise estimate and the posterior SNR based threshold adjuster serves to balance the effects of noise removal and speech preservation. In order to improve the final perceptual quality, a musical noise suppression algorithm and a TEO based silent segment smoothing module are also incorporated into the system. Simulation experiments are designed to demonstrate the capability of the proposed system.

A method for designing IIR digital filters with a novel stability criterion based on the argument principle is proposed in this paper. Unlike the stability criteria used in some design algorithms, this stability condition is both sufficient and necessary. In the paper, the weighted least-squares (WLS) design of IIR filters is first formulated as an iterative quadratic programming (QP) problem without any constraint. Then the stability criterion is incorporated in the quadratic form at each iteration. Two examples are presented to illustrate the effectiveness of the proposed approach.

In this paper, a peak-constrained weighted least-squares (PCWLS) method is presented for designing a general FIR digital filter with constant group delay passband, using the projected least-squares (PLS) algorithm. In the design of a general FIR filter, peak error constraints are nonlinear functions of filter coefficients and can not be accepted by the PLS algorithm. In the proposed method, peak error constraints are formulated, through some approximation, as a set of linear inequalities to be incorporated into the design problem. Design examples show the effectiveness of the peak error control of the proposed method.

In this paper, a novel method for designing IIR variable fractional delay (VFD) digital filters with variable and fixed denominator is presented. First of all, a peak-constrained weighted least-squares (PCWLS) method is employed to design a set of FIR fixed fractional delay (FD) filters according to given specifications. The PCWLS FIR filters are implemented by the projected least-squares (PLS) algorithm. An iterative WLS model reduction technique is utilized to design denominators, which can guarantee the stability of designed IIR VFD filter if the iteration converges. The numerator of IIR fixed FD filters can be designed by two approaches: The Approach 1 solves linear equations based on the orthogonality principle; and the Approach 2 formulates the numerator design problem as a standard quadratic programming (QP) problem. The coefficients of IIR fixed FD filters are finally approximated by polynomial functions of FD. Three sets of examples are given to demonstrate the effectiveness of the proposed method.

This paper presents an online procedure that produces the smallest feasible size of two-dimensional (2-D) FIR filters with prescribed magnitude error constraint. The procedure uses the mean square normalized error of constrained and unconstrained least-square filters to produce the initial and the subsequent sizes that converge to the smallest feasible one in a few iterations, where the constrained least-square filters are defined as the least-square filters satisfying the magnitude error constraint. The procedure finally returns a smallest size filter that satisfies the magnitude error constraint and has least total squared magnitude error. Design examples of diamond-shaped, rectangular, and elliptic filters are provided, and comparisons with an exhaustive search are given.

The Generalized Sidelobe Canceller (GSC) efficiently realizes the optimal Linearly Constrained Minimum Variance (LCMV) beamformer and delivers excellent beamforming results by reducing directional interference and noise. However, when the input signals are contaminated by other type of noises, such as background and diffused noises, the overall performance of GSC can be even worse than the traditional delay-and-sum beamformers. In this paper, we investigate the application of the spatially adaptive multiwavelet (MWT) denoising technique to the GSC in an environment with severe diffused noise. Comparing with the traditional scalar wavelets, the multiwavelets can better characterize the noise information in the signal such that better denoising performance can be achieved. Different approaches for integrating the GSC and the multiwavelet denoiser were studied. It is found that by adding two denoisers, one to the fixed constrained part and another to the final GSC output, an improvement of 2dB in SNR can be achieved as compared with the traditional GSC method.

Based upon discrete-time chaotic map and fractional delay of discrete-time signal, a new concept named fractional discrete-time chaotic map is introduced in this paper. Moreover, the ideal iterative function is changed into an approximate and realizable expression. As being evidence, fractional logistic discrete-time chaotic map is implemented, and its various richer chaotic behaviors are also shown

In this paper, a design approach for stable IIR digital filters with equiripple magnitude in passband and stopband and linear phase in passband is presented. The design approach adopts a constant group delay all-pole filter in the denominator and a mirror image polynomial transfer function in the numerator. Starting with a suitable filter order and a constant group delay value for the denominator, the Remez exchange algorithm is used to aid the design of the equiripple passband and stopband numerator transfer function. Equiripple passband and stopband characteristic offers benefits of reduced width in transition band and lower filter order. Lowpass, highpass, bandpass, and bandstop filter design examples are given.

A real discrete Gabor expansion (RDGE) for finite and infinite sequences is defined in this paper. By replacing the complex Gabor basis functions of the complex discrete Gabor expansion (CDGE) with real Gabor basis functions, a significant computation of the RDGE can be saved as compared with the computation of the CDGE. The similarity between the RDGE and the discrete Hartley transform (DHT) allows the RDGE to utilize the fast DHT algorithms for fast computation. In addition, the RDGE has a simple relationship with the CDGE such that the CDGE coefficients can be directly computed from the RDGE coefficients. Some simulation results are given at the end of this paper.

In this paper, fuzzy inference models for pattern classifications have been developed and fuzzy inference networks based on these models are proposed. Most of the existing fuzzy rule-based systems have difficulties in deriving inference rules and membership functions directly from training data. Rules and membership functions are obtained from experts. Some approaches use backpropagation (BP) type learning algorithms to learn the parameters of membership functions from training data. However, BP algorithms take a long time to converge and they require an advanced setting of the number of inference rules. The work to determine the number of inference rules demands lots of experiences from the designer. In this paper, self-organizing learning algorithms are proposed for the fuzzy inference networks. In the proposed learning algorithms, the number of inference rules and the membership functions in the inference rules will be automatically determined during the training procedure. The learning speed is fast. The proposed fuzzy inference network (FIN) classifiers possess both the structure and the learning ability of neural networks, and the fuzzy classification ability of fuzzy algorithms. Simulation results on fuzzy classification of two-dimensional data are presented and compared with those of the fuzzy ARTMAP. The proposed fuzzy inference networks perform better than the fuzzy ARTMAP and need less training samples.

A fuzzy inference network (FIN) is proposed. The proposed FIN preserves the advantages of both fuzzy classification algorithm and neural networks. It can learn membership functions directly from training samples and classify patterns according to the membership values. As efficient self-organizing learning algorithm is also presented.

A design algorithm for two-layer feedforward neural networks (2FNNs) for discrete input-output mapping is proposed. In this algorithm, uniformly quantized discrete weights are used, which could be in the form of either one-powers-of-two (OPOT) values or sum-of-powers-of-two (SPOT) values. The simplified sigmoid activation functions (SSAFs) are used at hidden neurons and the step functions are used at output neurons to further reduce the hardware implementation cost. Simulation results indicate that such networks can retain nearly identical recall performances as those of the corresponding networks using continuous weights and sigmoid activation functions (SAFs), while having increased computational speed in applications and reduced cost in digital hardware implementation.>

Defines four types of fuzzy neurons and proposes the structure of a four-layer feedforward fuzzy neural network (FNN) and its associated learning algorithm. The proposed four-layer FNN performs well when used to recognize shifted and distorted training patterns. When an input pattern is provided, the network first fuzzifies this pattern and then computes the similarities of this pattern to all of the learned patterns. The network then reaches a conclusion by selecting the learned pattern with the highest similarity and gives a nonfuzzy output. The 26 English alphabets and the 10 Arabic numerals, each represented by 16/spl times/16 pixels, were used as original training patterns. In the simulation experiments, the original 36 exemplar patterns were shifted in eight directions by 1 pixel (6.25% to 8.84%) and 2 pixels (12.5% to 17.68%). After the FNN has been trained by the 36 exemplar patterns, the FNN can recall all of the learned patterns with 100% recognition rate. It can also recognize patterns shifted by 1 pixel in eight directions with 100% recognition rate and patterns shifted by 2 pixels in eight directions with an average recognition rate of 92.01%. After the FNN has been trained by the 36 exemplar patterns and 72 shifted patterns, it can recognize patterns shifted by 1 pixel with 100% recognition rate and patterns shifted by 2 pixels with an average recognition rate of 98.61%. The authors have also tested the FNN with 10 kinds of distorted patterns for each of the 36 exemplars. The FNN can recognize all of the distorted patterns with 100% recognition rate. The proposed FNN can also be adapted for applications in some other pattern recognition problems.>

A model for artificial neural networks is introduced. The model consists of a multiple layers of identical feedbackward structures. Both learning and recall algorithms of the model are presented, and simulation results are presented.>

A normalized digital lattice structure is presented for the realization of 1-D ARMA (autoregressive moving average) digital filters. The resultant realization structure is minimal in terms of the numbers of parameters and delays while possessing the desirable properties of the classical AR lattice filters. A fourth-order elliptic filter was realized using the procedures presented.>

A novel implementation of the counterpropagation network with both the recalling and the learning algorithms is presented. The implementation is based on a word-level systolic architecture which consists of two types of basic cells. The final architecture is characterized by regular interconnections, which is very suitable for VLSI implementation.>

Two learning algorithms for storing data in an autoassociator with full storage capacity are presented. The first algorithm can teach a neural network to store short-bit-length data with full storage capacity (as compared to the Hebb rule), and the second algorithm can do the same for long-bit-length data at a higher speed (as compared to the delta rule). The algorithms eliminate the problem of limited storage capacity associated with the Hebb rule, and their learning speeds are much faster than that of the corresponding delta rule. Simulations have been done to compare the two learning algorithms to those of the Hebb rule and the delta rule. Results show that the two algorithms are much better than the Hebb rule in terms of storage capacity and much better than the delta rule in terms of learning speed

In this paper, a delayed multipath method for the realization of two-dimensional nonrecursive and recursive digital filters is presented. In this method, a two-dimensional transfer function polynomial is decomposed into a number of shorter transfer function polynomials in parallel. Each of these transfer function polynomials can be realized efficiently in terms of a number of processors. Consequently, high speed computation can be obtained which can be applied to various digital filtering applications. For a two-dimensional finite impulse response digital filter of orders M1 and M2, the maximum throughput can be (M2+1)[(M1+1)/2]2times that of the conventional direct form realization using one processor.

The theory of quarter-plane autoregressive (AR) lattice parameter modeling of two-dimensional (2-D) fields is extended to three-dimensional (3-D) case with cubic-space support involving seven lattice parameters and eight prediction error fields at each stage. The relationship between the cubic-space 3-D transfer function and lattice parameters, and the sufficient lattice parameter stability conditions are also presented. Numerical examples are included to verify the formulation.

An outline for the design of two-dimensional passive state-space digital filters in first-order form and with the possibility of eliminating zero-input and constant-input oscillations is presented. The method is applicable to the design of all types of frequency-band filters and is suitable for circuit integration.

A linear transformation approach to the design of a family of passive second-order digital filters is presented. Starting from the definition of the power dissipation in a two-port network, a power dissipation expression for a linear transformed two-port network is derived. A passive recursive second-order structure can be obtained by linear transformation of a passive gyrator circuit. By trying all feasible Voltage, Current, Power, Voltage -current, Current-voltage transformations, twenty-one new passive second-order digital filter structures were obtained. Under finite arithmetic operations, this family of digital filters is free from zero-input and forced response oscillations by applying magnitude truncation at delays' inputs.