Search

This paper explores the design and hardware efficient realization of low complexity filter bank-based reconfigurable channelizers suitable for wireless receivers. The reconfigurability is achieved by deriving a set of non-uniform filter banks with different sampling factor combinations from a low complexity partial cosine modulated uniform filter bank, by combining different number of its neighbouring channels. The hardware complexity comparison of the filter bank channelizer designed by the proposed approach, with the existing works proves that this approach requires less number of multipliers in its realization to extract the same set of communication standards. To reduce the power and complexity of the multipliers needed for the hardware realization of the filter coefficients, an approximation algorithm, which increases the zeros in the binary represented filter coefficients is employed. The synthesis results show that the algorithm offers a considerable reduction in the hardware and power consumptions of the filter bank. [ABSTRACT FROM AUTHOR]

In this paper, a low-energy switching scheme based on the asymmetric binary search algorithm for successive approximation register (SAR) analog-to-digital converters (ADCs) is proposed. To reduce switching energy and area, the proposed scheme uses the single-side switching method and the least significant bit (LSB) capacitor is split into two one-half unit capacitors. Using the series capacitance scheme, the capacitance mismatch caused by process deviation is effectively improved. Compared with the conventional switching method, the switching energy of this scheme is decreased by 99.43% and the area is saved by 87.2%. MATLAB simulation results show that this scheme also has a low nonlinearity that the maximum differential nonlinearity is only 0.37 LSB and the maximum integral nonlinearity is only 0.43 LSB. [ABSTRACT FROM AUTHOR]

We provide here an optimal method of approximating a signal by piecewise constant functions. To this end, we minimize over the signal subdomains a fidelity term between the signal and its corresponding piecewise approximations; subdomains being determined by the number of approximations samples used for. An optimal recursive relationship is then obtained and proven, which helps us to derive the proposed approximation algorithm. The complexity of the algorithm is O( MN 2 ), where N is the number of samples of the processed signal and M is the number of piecewise constant approximation functions. There are different techniques to approximate a signal using piecewise constant functions, wavelet decomposition is one of them by means of a Haar wavelet. Our approach is then compared to linear and nonlinear wavelet-based approximations, and both qualitative and quantitative results are provided on various tested signals, showing the efficiency of the proposed approach. [ABSTRACT FROM AUTHOR]

This paper deals with order reduction in discrete-time systems. The reduction technique is based on utilization of Routh approximation and multiplicative operator. Various combinations of Routh table are used to derive the desired numerator and denominator polynomial of reduced-order model. Numerical examples are presented to verify the proposed algorithm. [ABSTRACT FROM AUTHOR]

In this paper, a new technique for weak signal detection using stochastic resonance (SR) with approximated fractional integrator (AFI) has been investigated. To improve the performance of weak signal detector, AFI is convolved with noisy samples which (i.e., AFI) attenuates the noise because of its low-pass filtering nature. SR noise ( i.e., some fixed amount of noise) is added externally, which causes further improvement in detection. The external noise under SR has been obtained in Neyman-Pearson (NP) framework. For comparison, receiver operating characteristic curve has been analysed. The parameters, probability of detection (PD) and deflection coefficient ratio are also calculated at a fixed value of probability of false alarm (PFA). It has been observed that the performance of the proposed detector is better or comparable to most of the state-of-the-art techniques. [ABSTRACT FROM AUTHOR]

This study is conducted to investigate stability and control problems within a finite-time interval for a 2D continuous-discrete system in Roesser model. The concepts of finite-time stability (FTS) and finite-time boundedness (FTB) are naturally extended to the 2D continuous-discrete system. Recursive relations between system states are first obtained, then sufficient conditions for FTS and FTB in the system are derived, and a finite-time controller is supplied to the system. Sufficient conditions for finite-time stabilization are also provided for the linear repetitive process. Examples of metal rolling operation are presented to illustrate the proposed method. [ABSTRACT FROM AUTHOR]

Online arithmetic operators offer advantages of reduction in resource utilization and interconnection complexity besides providing pipelining at digit level. Multiplierless constant coefficient multiplication using the shift-and-add technique is widely used in digital signal processing applications. This paper proposes a novel bit serial adaptation of the parallel shift-and-add algorithm to online arithmetic. The proposed multipliers use right shifts instead of the traditional left shifts resulting in causal online implementations. Graph-based and hybrid algorithms are developed for the estimation of the distance of a constant from a set of constants in terms of the number of additions and for the synthesis of online multiple constant multipliers under area and online delay constraints. The computational complexity of the algorithms is determined. Results of implementation on randomly generated constant sets and FIR filter instances show substantial improvements in the number of operations required using the distance heuristic. Further, it is shown that the proposed techniques and algorithms result in significant savings in resource utilization, logic depth, and clock frequency compared to parallel and digit-serial algorithms. [ABSTRACT FROM AUTHOR]

Formants (peaks in the spectrum) are typically found in vowel regions. However, resonances of vocal tract are present in both voiced and unvoiced regions. It is important to focus on extracting vocal tract resonances from speech signal to provide compact and effective representation. This work applies multilayered neural networks for the estimation of vocal tract resonances from linear prediction cepstral coefficients. To train such a neural network, the voiced regions of speech signal are first probed for reliable spectral prominent regions. A sequence of spectral peaks detected by imposing continuity constraints in voiced regions is referred to as reliable spectral prominent regions. A multilayer feedforward neural network is then trained to model a mapping function from linear prediction cepstral coefficients to vocal tract resonances estimated in these reliable spectral prominent regions. This modeled function is used to estimate vocal tract resonances in all speech regions including unvoiced sections. Improvements in vocal tract resonances estimated by this modeled function are 37, 35 and 41%, respectively, for the first three resonances when compared to WaveSurfer, and 49, 32 and 29% for the first three resonances when compared to PRAAT. [ABSTRACT FROM AUTHOR]

The robust finite-time H∞ control for a kind of uncertain delay systems with large delay period (LDP) is discussed in this paper. First, a switching technique is exploited to transform the original system into a switched delay system. Second, within the limitation of frequency and length rate of LDP, a state feedback controller is designed to guarantee that the closed-loop system is robust finite-time bounded. Third, the finite-time H∞ performance analysis for the closed-loop system is developed. Finally, two examples are presented to clarify the validity of the proposed approach. [ABSTRACT FROM AUTHOR]

This article presents the performance evaluation of Vector Fitting Algorithm (VFA) from a system identification perspective. In this paper, VFA has been first applied to known baseband and bandpass systems such as Butterworth lowpass and bandpass filters to analyze the algorithm’s pole-residue extraction ability for band-limited noisy data. The poles identified by the algorithm for different bandwidths and noise powers are compared with the actual system poles of the baseband and bandpass systems. It is concluded that the algorithm is capable of identifying the actual system poles even if the capture bandwidth is less than the 3 dB bandwidth, which is a significant observation of this paper. It is also seen that the system identification performance with noisy data is better for baseband systems when compared to bandpass systems. Further, a practical investigation has been done to evaluate VFA performance for modeling a microstrip coupled line filter in the presence of noise. [ABSTRACT FROM AUTHOR]

Deep learning-based speech enhancement approaches like deep neural networks (DNN) and Long Short-Term Memory (LSTM) have already demonstrated superior results to classical methods. However, these methods do not take full advantage of temporal context information. While DNN and LSTM consider temporal context in the noisy source speech, it does not do so for the estimated clean speech. Both DNN and LSTM also have a tendency to over-smooth spectra, which causes the enhanced speech to sound muffled. This paper proposes a novel architecture to address both issues, which we term a conditional generative model (CGM). By adopting an adversarial training scheme applied to a generator of deep dilated convolutional layers, CGM is designed to model the joint and symmetric conditions of both noisy and estimated clean spectra. We evaluate CGM against both DNN and LSTM in terms of Perceptual Evaluation of Speech Quality (PESQ) and Short-Time Objective Intelligibility (STOI) on TIMIT sentences corrupted by ITU-T P.501 and NOISEX-92 noise in a range of matched and mismatched noise conditions. Results show that both the CGM architecture and the adversarial training mechanism lead to better PESQ and STOI in all tested noise conditions. In addition to yielding significant improvements in PESQ and STOI, CGM and adversarial training both mitigate against over-smoothing. [ABSTRACT FROM AUTHOR]

In this paper, a novel diffusion estimation algorithm is proposed by combining the adapt-then-combine (ATC) diffusion NLMS (DNLMS) and the sign operation to the error signals at all agents. Due to using the sign operation, the proposed algorithm has a robust performance against impulsive noise. A variable step size is obtained by minimizing the mean-square deviation to achieve fast convergence rate and small steady-state error. Meanwhile, the proposed algorithm has a good tracking capability when the system suddenly changes. Simulations on system identifications demonstrate that the proposed diffusion algorithm shows a good performance under an impulsive noise scenario. [ABSTRACT FROM AUTHOR]

This paper introduces the idea of versatile circuits for multiplying 4-dimensional hypercomplex numbers in hardware. Depending on the settings of such a device, a variable quaternion can be left- or right-multiplied by a constant coefficient or by its conjugate, as various operations are useful in transform-type algorithms. Multiplierless circuits based on distributed arithmetic (DA) are reviewed that compute quaternion products by additions and bit shifts. It is shown that they can be made versatile by extending memory of partial results, but a better solution is our method for preprocessing bits used to address this memory. The method allows for using the same partial results to compute different inner products that are related to a quaternion multiplication. So versatile multipliers can be implemented with memory optimized so as to save area or to speed up reprogramming compared to the basic DA-based circuit. This has been demonstrated by hardware design experiments, which show that 13-69% area can be saved in a case of ASIC implementation, while for FPGA implementation, spending only 11% more logic resources allows a multiplier to be reprogrammed 75% faster. Additionally, it has been explained how versatile multipliers can be used to realize low-area analysis/synthesis filter banks. [ABSTRACT FROM AUTHOR]

The three-dimensional video (3DV) is composed of variable-length stream sequences captured via diversified cameras surrounding an object. Thus, it is an urgent task to accomplish sufficient encoding to be compatible with incoming bandwidth demands, while achieving a recommended 3DV reception performance. In the 3DV compression framework, the lost macro-blocks (MBs) might propagate into the following frames and the adjoining views. Therefore, it is obligatory to avoid error propagation by concealing the corrupted MBs at the decoder through the utilization of appropriate post-processing error concealment (EC) techniques. The existing EC algorithms fundamentally exploit the temporal, inter-view, and spatial matching within the 3DV frames and views to reconstruct the disparity vectors (DVs) and motion vectors (MVs) of the corrupted MBs. Unluckily, in the state of high severe corruptions and heavily erroneous MBs, these concealment algorithms are predominantly unreliable and might give unreliable 3DV quality. Thence, in this work, we suggest the utilization of the outer block boundary matching algorithm to estimate the MVs and the directional interpolation EC algorithm to estimate the DVs of the erroneous MBs. After that, the Bayesian Kalman filter (BKF) is employed because of its efficiency to filter out the inherent errors in the previously predicted DVs and MVs to accomplish better 3D video performance. Experimental results on standard 3DV sequences demonstrate that the suggested BKF-based EC scheme is more powerful with heavy losses. It subjectively and objectively outperforms the traditional concealment techniques at severely random and bursty packet loss rates (PLRs). [ABSTRACT FROM AUTHOR]

This paper addresses the problem of estimating the parameters of adaptive directional time-frequency distributions (ADTFDs). ADTFDs locally optimize the direction of the smoothing kernel on the basis of directional Gaussian or double derivative directional Gaussian filter. Conventionally, the parameters of these techniques have to be tuned manually for each particular signal. Global optimization of the parameters fails to provide the desired results when the signal encompasses close or short-duration components. We propose a two-stage algorithm to resolve this issue. As part of the first stage, the length of the smoothing kernel is optimized globally. In the second stage, the parameters which control the shape of the selected smoothing window are optimized, locally. It is shown that the multistage algorithm can result in a time-frequency distribution that has both high resolution for close components and good concentration of signal energy for short-duration signal components. Experimental findings reveal the superiority of the proposed technique over the existing methods in the case of complete signals and its benefits in the case of signals with missing samples. [ABSTRACT FROM AUTHOR]

Recognizing children’s speech on automatic speech recognition (ASR) systems developed using adults’ speech is a very challenging task. As reported by several earlier works, a severely degraded recognition performance is observed in such ASR tasks. This is mainly due to the gross mismatch in the acoustic and linguistic attributes between those two groups of speakers. One among the various identified sources of mismatch is that the vocal organs of the adult and child speakers are of significantly different dimensions. Feature-space normalization techniques are noted to effectively address the ill-effects arising from those differences. Two most commonly used approaches are the vocal-tract length normalization and the feature-space maximum-likelihood linear regression. Another important mismatch factor is the large variation in the average pitch values across the adult and child speakers. Addressing the ill-effects introduced by the pitch differences is the primary focus of the presented study. In this regard, we have explored the feasibility of explicitly changing the pitch of the children’s speech so that observed pitch differences between the two groups of speaker are reduced. In general, speech data from children is high-pitched in comparison with that from the adults’. Consequently, in this study, the pitch of the adults’ speech used for training the ASR system is kept unchanged while that for the children’s test speech data is reduced. Significant improvement in the recognition performance is noted by this explicit reduction of pitch. To conserve the critical spectral information and to avoid introducing perceptual artifacts, we have exploited timescale modification techniques for explicit pitch mapping. Furthermore, we also presented two schemes to automatically determine the factor by which the pitch of the given test data should be varied. Automatically determining the compensation factor is critical since an ASR system is expected to be accessed by both adult and child speakers. The effectiveness of proposed techniques is evaluated on adult data trained ASR systems employing different acoustic modeling approaches, viz. Gaussian mixture modeling (GMM), subspace GMM and deep neural networks (DNN). The proposed techniques are found to be highly effective in all the explored modeling paradigms. To further study the effectiveness of the proposed approaches, another DNN-based ASR system is developed on a mix of speech data from adult as well as child speakers. The use of pitch reduction is observed to be effective even in this case. [ABSTRACT FROM AUTHOR]

Face recognition becomes a challenging topic in several fields since images of faces are varied by changing illuminations, facial rotations, facial expressions, etc. In this paper, two dimensional discrete multiwavelet transform (2D DMWT) and fast independent component analysis (FastICA) are proposed for face recognition. Preprocessing, feature extraction, and classification are the main steps in the proposed system. In the preprocessing step, each pose in the database is divided into six parts to reduce the effect of unnecessary facial features and highlight the local features in each part. For feature extraction, the 2D DMWT is applied to each part for dimensionality reduction and features extraction. This results in two facial representations. Then FastICA followed by ℓ2-norm is applied to each representation, which produces six and three different techniques for the first and second representation, respectively. This results in features that are more discriminating, less dependent, and more compressed. In the recognition step, the resulted compressed features from the two representations are fed to a neural network-based classifier for training and testing. The proposed techniques are extensively evaluated using five databases, namely ORL, YALE, FERET, FEI, and LFW, which have different facial variations, such as illuminations, rotations, facial expressions, etc. The results are analyzed using K-fold cross-validation. Sample results and comparison with a large number of recently proposed approaches are provided. The proposed approach is shown to yield significant improvement compared with the other approaches. [ABSTRACT FROM AUTHOR]

A new kind of filter called Legendre chained-function (LCF) filter with characteristic function given by the product of low-degree Legendre orthogonal polynomials (seed functions) is studied in this paper. LCF filter magnitude response exhibits unequal ripple level in the passband compared to Chebyshev chained-function filter with identical edge ripple factor at the passband edge. A proper combination of seed functions, i.e., a product of them, is used to control the maximum ripple in the passband, which affects the return loss, selectivity and a group delay characteristics of a filter. By selecting one of the possible combinations of seed functions (thus obtaining various degrees of freedom in filter design), filters with improved performances compared to traditional approximation techniques can be obtained. The degrees of freedom increase if the degree of filter increases. Compared to existing Chebyshev chained-function (CCF) filters, whose performances are also presented, and Butterworth function filters as a special case of both LCF and CCF filters, the new family of LCF filters has many advantages. A table summarizing properties of CCF and LCF filters is given for design purpose. [ABSTRACT FROM AUTHOR]

Nowadays, portable multimedia devices have been widely used in low-power structures. Additionally, using approximate computing and multiple-valued logic (MVL) can lead to lower power consumption. Carbon nanotube field effect transistor (CNTFET), with a determinable threshold voltage, is an appropriate choice for designing MVL circuits. In this study, a new approximate ternary full adder is presented based on CNTFET switching logic with the aim of reducing area and energy consumption. The results obtained by the Synopsys HSPICE simulator with the Stanford 32 nm CNTFET technology show that the average power consumption, delay, and energy consumption of the proposed designs are significantly lower than the other conventional and state-of-the-art CNTFET-based ternary circuits. In addition, the proposed ternary full adder is applied to reduce the steps of a ternary multiplier structure and some important metrics such as normalized error distance and energy-error trade-off are calculated to evaluate the efficiency of the proposed circuits. The layout of the proposed design is also plotted to estimate area consumption. Moreover, the approximate design is used for image processing applications such as image multiplying and image blending and the results are accordingly compared with the exact outputs. Similarly, the evaluated peak signal-to-noise ratio and mean structural similarity index metric indicate that the proposed approximate full adder has a proper accuracy for image processing applications. [ABSTRACT FROM AUTHOR]

A new modified pseudo-Taylor exponential approximation is presented for realizing a current-to-current exponential function generator. The proposed approximation for exponential function generation has been implemented using translinear principle of MOSFETs operating in weak inversion region. The SPECTRE simulation tool from Cadence, with 180 nM CMOS process parameters, has been utilized for testing the workability of the CMOS implementation of the proposed approximation, which uses only $$\pm 0.5\hbox {V}$$ power supply. The post-layout simulation results of the proposed CMOS exponential generator, thus, consume only $${\approx }119\,\hbox {nW}$$ power and produce 51.6 dB range of linear-in-dB output with only $$\pm 0.21\,\hbox {dB}$$ error while occupying an area of $$0.26\,\upmu \hbox {m}^{2}$$ . [ABSTRACT FROM AUTHOR]

A novel adaptive step-size matching pursuit algorithm (AStMP) is proposed. AStMP reduces the computational cost and increases the accuracy of reconstructing practical signals (i) whose sparsity K is unknown and/or (ii) may be corrupted by noise. AStMP accurately estimates K by combining the sparsity estimate of sparsity adaptive subspace pursuit (SASP) with the adaptively changing stepsize at each stage of sparsity adaptive matching pursuit (SAMP). Thus, AStMP can quickly achieve accurate estimation of the sparsity level and the true support set of the target signals. Meanwhile, a preselection is employed to reduce the computational complexity of each stage. When K is greater than half the number of measurements M, the probability of exact recovery is improved by analyzing the support set. Since the stepsize changes adaptively, AStMP can be applied in both the noiseless and noisy cases when the signal is not strictly sparse, allowing for exact or approximate signal recovery. Experimental results demonstrate that the AStMP is effective in fast and exact reconstruction and has good performance. [ABSTRACT FROM AUTHOR]

This paper describes the distributed information filtering where a set of sensor networks are required to simultaneously estimate input and state of a linear discrete-time system from collaborative manner. Our research purpose is to develop a consensus strategy in which sensor nodes communicate within the network through a sequence of Kalman iterations and data diffusion. A novel recursive information filtering is proposed by integrating input estimation error into measurement data and weighted information matrices. On the fusing process, local system state filtering transmits estimation information using the consensus averaging algorithm, which penalizes the disagreement in a dynamic manner. A simulation example is provided to compare the performance of the distributed information filtering with optimal Gillijins-De Moor's algorithm. [ABSTRACT FROM AUTHOR]

The alternation theorem is the core of efficient approximation algorithms for the minimax design of finite-impulse response (FIR) filters. In this paper, an extended alternation theorem with additional mixed constraints, i.e., equality-and-inequality constraints, is obtained. Then, an efficient multiple-exchange algorithm based on the extended theorem is presented for designing linear-phase FIR filters with frequency mixed constraints in the minimax sense. Further, convergence of the algorithm is established. Several design examples and comparisons with existing techniques are presented, and the simulation results show that the proposed algorithm is numerically more efficient and guaranteed to converge to the optimal solution. [ABSTRACT FROM AUTHOR]

The original version of the article is missing the below funding information. [ABSTRACT FROM AUTHOR]