We construct a fast method, O(N log N), for the computation of discrete Gauss transforms with complex parameters, capable of dealing with unequally spaced grid points. The method is based on Fourier techniques, and in...
详细信息
We construct a fast method, O(N log N), for the computation of discrete Gauss transforms with complex parameters, capable of dealing with unequally spaced grid points. The method is based on Fourier techniques, and in particular it makes use of a modified unequally spaced fast Fourier transform algorithm, in combination with previously suggested divide and conquer strategies for ordinary fast Gauss transform methods. (C) 2004 Elsevier Inc. All rights reserved.
A new fast and robust 3-level binary higher order statistics (HOS) based algorithm for simultaneous voiced/unvoiced detection and pitch estimation of speech signals in coloured noise environments with low SNR is prese...
详细信息
A new fast and robust 3-level binary higher order statistics (HOS) based algorithm for simultaneous voiced/unvoiced detection and pitch estimation of speech signals in coloured noise environments with low SNR is presented. The use of the 3-level binary speech signals dramatically reduces the computational effort required in evaluating the higher order cumulants. The superior performance of the new algorithm to the conventional autocorrelation method using real speech signals is demonstrated. The algorithm can easily be implemented in digital hardware using simple combinatorial logic. (C) 1997 Elsevier Science B.V.
A new type of coded aperture configuration that enables fast decoding of the coded aperture shadowgram data is presented. Based on the products of incidence vectors generated from the Singer difference sets, we call t...
详细信息
A new type of coded aperture configuration that enables fast decoding of the coded aperture shadowgram data is presented. Based on the products of incidence vectors generated from the Singer difference sets, we call these Singer product apertures. For a range of aperture dimensions, we compare experimentally the performance of three decoding methods: standard decoding, induction decoding and direct vector decoding. In all cases the induction and direct vector methods are several orders of magnitude faster than the standard method, with direct vector decoding being significantly faster than induction decoding. For apertures of the same dimensions the increase in speed offered by direct vector decoding over induction decoding is better for lower throughput apertures.
作者：
Boag, ALetrou, CTel Aviv Univ
Dept Elect Engn Phys Elect IL-69978 Tel Aviv Israel CNRS
FRE 2310 GETINT Lab SAMOVAR F-91011 Evry France
A novel algorithm referred to as the fast physical optics (FPO) for computing the radiation patterns of nonplanar aperture antennas over a range of observation angles is presented. The computation is performed in the ...
详细信息
A novel algorithm referred to as the fast physical optics (FPO) for computing the radiation patterns of nonplanar aperture antennas over a range of observation angles is presented. The computation is performed in the framework of the conventional physical optics approximation appropriate for the high frequency regime. The proposed algorithm is directly applicable to reflector and lens antennas as well as to radomes. The method comprises two steps. First, a decomposition of the aperture into subdomains and computation of the pertinent radiation pattern of each subdomain. Second, interpolation, phase-correction and aggregation of the radiation patterns into the final pattern of the whole aperture. A multilevel algorithm is formulated via a recursive application of the domain decomposition and aggregation steps. The computational structure of the multilevel algorithm resembles that of the FFT while avoiding its limitations.
Cross-validation (CV) is a very popular technique for model selection and model validation. The general procedure of leaveone-out CV (LOO-CV) is to exclude one observation from the data set, to construct the fit of th...
详细信息
Cross-validation (CV) is a very popular technique for model selection and model validation. The general procedure of leaveone-out CV (LOO-CV) is to exclude one observation from the data set, to construct the fit of the remaining observations and to evaluate that fit on the item that was left out. In classical procedures such as least-squares regression or kernel density estimation, easy formulas can be derived to compute this CV fit or the residuals of the removed observations. However, when high-breakdown resampling algorithms are used, it is no longer possible to derive such closed-form expressions. High-breakdown methods are developed to obtain estimates that can withstand the effects of outlying observations. fast algorithms are presented for LOO-CV when using a high-breakdown method based on resampling, in the context of robust covariance estimation by means of the MCD estimator and robust principal component analysis. A robust PRESS curve is introduced as an exploratory tool to select the number of principal components. Simulation results and applications on real data show the accuracy and the gain in computation time of these fast CV algorithms. (c) 2006 Elsevier B.V. All rights reserved.
In this work, we study the computational aspects of the algorithms for testing the strict sense and wide sense stability of discrete time systems. These algorithms are based on the computation of the reflection coeffi...
详细信息
In this work, we study the computational aspects of the algorithms for testing the strict sense and wide sense stability of discrete time systems. These algorithms are based on the computation of the reflection coefficients for a given real polynomial. The algorithms presented here require one-half the number of multiplications and the same number of additions as compared to the previously known similar algorithms. In dieser Arbeit werden Algorithmen untersucht, welche die Stabilität im weiteren und engeren Sinne für zeitdiskrete Systeme prüfen. Diese Algorithmen basieren auf der Bestimmung der Reflexionskoeffizienten für ein gegebenes reelles Polynom. Die hier vorgestellten Verfahren benötigen bei gleicher Anzahl von Additionen nur halb so viele Multiplikationen wie die bisher bekannten vergleichbaren Stabilitätstests. On examine les particularités numériques des algorithmes proposés pour tester la stabilité au sens large, ainsi que celle au sens strict des systèmes linéaires à temps discret. Ces algorithmes sont basés sur la récursion inverse de Levinson au cours de laquelle on calcule les coefficients de corrélation partielle associés à un polynôme réel. Dans leur nouvelle formulation, ils exigent environ un nombre de multiplications moitié moindre que dans la formulation classique, tout en préservant le nombre d'additions.
A new two-dimensional (2D) Walsh transform scheme is presented based on a block splitting technique. The advantage of the proposed approach lies in the simple data flow architecture and the reduction in array accessin...
详细信息
A new two-dimensional (2D) Walsh transform scheme is presented based on a block splitting technique. The advantage of the proposed approach lies in the simple data flow architecture and the reduction in array accessing operations compared with the conventional row-column (RC) scheme. The new scheme exploits adequately two local accessing properties of the cache: the time local property and the space local property. The new 2D scheme is faster than the RC scheme when applying the same one-dimensional (ID) algorithm. (c) 2006 Elsevier B.V. All rights reserved.
Subspace methods have proven to be efficient for the identification of linear time-invariant systems, especially applied to mechanical, civil or aeronautical structures in operation conditions. Therein, system identif...
详细信息
Subspace methods have proven to be efficient for the identification of linear time-invariant systems, especially applied to mechanical, civil or aeronautical structures in operation conditions. Therein, system identification results are needed at multiple (over-specified) model orders in order to distinguish the true structural modes from spurious modes using the so-called stabilization diagrams. In this paper, new efficient algorithms are derived for this multi-order system identification with subspace-based identification algorithms and the closely related Eigensystem Realization Algorithm. It is shown that the new algorithms are significantly faster than the conventional algorithms in use. They are demonstrated on the system identification of a large-scale civil structure. (C) 2012 Elsevier Ltd. All rights reserved.
To solve the long training time in Waibel's time-delay neural networks (TDNNs) for phoneme recognition, several improved fast learning methods are put forward: (1) by a combination between the unsupervised Oja'...
详细信息
To solve the long training time in Waibel's time-delay neural networks (TDNNs) for phoneme recognition, several improved fast learning methods are put forward: (1) by a combination between the unsupervised Oja's rule learning method with the similar error backpropagation (BP) algorithm for initial weights training;(2) by improving of the error energy function with weights update according to the output error size;(3) by changing of BP error from along layers to frames and using the averaged overlapping part of frame-shift delta values in the weights of the bottom layer;(4) by training of the data from a small to large number of samples gradually;and (5) by using the optimal modular neural networks (OMNNs) with tree structure for multiple phonemes. Our experimental results indicate that the convergence speed is accelerated with orders of magnitude and in most cases the error function descends monotonically while the network complexity increases less and the recognition rates are almost the same among the different comparative experiments. (C) 2002 Elsevier Science Inc. All rights reserved.
In this brief, two different approaches to compute the 2-D discrete Hadamard transform (DHAT) of sizes 2(r) x 2(r), r > 1 are presented. The both are based on the so-caned 1-D discrete paired transforms, which spli...
详细信息
In this brief, two different approaches to compute the 2-D discrete Hadamard transform (DHAT) of sizes 2(r) x 2(r), r > 1 are presented. The both are based on the so-caned 1-D discrete paired transforms, which split the 2(r)-point DHAT into a set of smaller 2(r-z)-point transforms, i = 1 : r. The first approach is a column(rom)-wise algorithm which requires, on average, no more than 12 operations of additions per sample. The second approach is a splitting algorithm, in which the 2(r) x 2(r)-point DHAT is reduced to the computation of 2(r-1)3 1-D DHATs instead of 2(r+1) 1-D DHAT's in the column(rom)-wise algorithm, but requires more operations of additions. Examples and comparison of the proposed algorithms with the known algorithms are provided.
暂无评论