A fast algorithm for ∈R-1 factorization of Toeplitz matrices

نویسنده

  • George-Othon Glentis
چکیده

In this paper a new order recursive algorithm for the efficient 9B-l factorization of Toeplitz matrices is described. The proposed algorithm can be seen as a fast modified Gram-Schmidt method which recursively computes the orthonormal columns si, i = 1,2, . ,p, of 1, as well as the elements of WI, of a Toeplitz matrix X with dimensions L x p. The 2 factor estimation requires 8Lp MADS (multiplications and divisions). Matrix 9-l is subsequently estimated using 3p2 MADS. A faster algorithm, based on a mixed 9 and W’ updating scheme, is also derived. It requires 7Lp + 3.5~~ MADS. The algorithm can be efficiently applied to batch least squares FIR filtering and system identification. When determination of the optimal filter is the desired task it can be utilized to compute the least squares filter in an order recursive way. The algorithm operates directly on the experimental data, overcoming the need for covariance estimates. An orthogonalized version of the proposed _?2W-’ algorithm is derived. Matlab code implementing the algorithm is also supplied. Zusammenfassung In dieser Arbeit wird ein neuer ordnungsrekursiver Algorithmus zur effizienten Z!g’ Faktorisierung von Toeplitz-Matrizen beschrieben. Der vorgeschlagene Algorithmus kann als schnelle Modifizierte Gram-Schmidt-Methode betrachtet werden, bei der die orthonormalen Spalten di, i = 1,2, . . . ,p, von 9 sowie die Elemente von W’ einer Toeplitz-Matrix .% mit der Dimension L x p berechnet werden. Der 9-Faktor wird mit einem Aufw&d von 8Lp MADs (Multiplikationen und Divisionen) geschgtzt. Die Matrix W’ wird fortlaufend unter Anwendung von 3p2 MADs geschltzt. Weiterhin wird ein schnellerer Algorithmus abgeleitet, der auf einem gemischten 9 und W’ Aktualisierungsschema basiert. Er erfordert 7Lp + 3.5~~ MADs. Der Algorithmus kann effizient zur Echtzeit Least-Squares FIRFilterung und Systemidentifkation benutzt werden. Wenn die Bestimmung des Optimalfilters verlangt ist, kann er zur Berechnung des Least-Squares-Filter in ordnungsrekursiver Form herangezogen werden. Der Algorithmus arbeitet direkt mit Hilfe der experimentellen Daten und vermeidet die Schiitzung der Kovarianz. Es wird eine orthogonalisierte Version des vorgeschlagenen 499’ Algorithmus abgeleitet. Eine Matlab Implementierung des Algorithmus wird angegeben. Rbumiz Dans cet article, un nouvel algorithme rtcursif d’ordre permettant une factorisation 1@’ efficace des matrices de Toeplitz est dCcrit. L’algorithme propost: peut &tre vu comme une mCthode rapide de Gram-Schmidt Modifit: laquelle calcule rCcursivement les colonnes orthonormales li, i = 1,2, , p, de 9 ainsi que les iltments de W’ d’une matrice X de Toeplitz de dimension L x p. Le facteur 9 est estimC en terme de 8Lp MADS (Multiplications et Divisions). La matrice W ’ est estimte ultkrieurement en utilisant le MADS 3~‘. Un algorithme plus rapide basi: sur un schCma de rafraichissement mixte de 9 et de W-’ est Cgalement dtcrit. I1 nCcessite 7Lp + 3.5~~ MADS. L’algorithme peut Ctre 0165-1684,/95/$9.50

برای دانلود متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

ثبت نام

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

منابع مشابه

Fast and Stable Pascal Matrix Algorithms

In this paper, we derive a family of fast and stable algorithms for multiplying and inverting n × n Pascal matrices that run in O(n log n) time and are closely related to De Casteljau’s algorithm for Bézier curve evaluation. These algorithms use a recursive factorization of the triangular Pascal matrices and improve upon the cripplingly unstable O(n log n) fast Fourier transform-based algorithm...

متن کامل

Transformation Techniques for Toeplitz and Toeplitz-plus-hankel Matrices Part Ii. Algorithms

In the rst part 13] of the paper transformationsmappingToeplitz and Toeplitz-plus-Hankel matrices into generalizedCauchy matrices were studied. In this second part fast algorithms for LU-factorization and inversion of generalized Cauchy matrices are discussed. It is shown that the combinationof transformation pivoting techniques leads to algorithms for indeenite Toeplitz and Toeplitz-plus-Hanke...

متن کامل

NORTH- HOLLAND High Performance Algorithms for Toeplitz and Block Toeplitz Matrices

In this paper, we present several high performance variants of the classical Schur algorithm to factor various Toeplitz matrices. For positive definite block Toeplitz matrices, we show how hyperbolic Householder transformations may be blocked to yield a block Schur algorithm. This algorithm uses BLAS3 primitives and makes efficient use of a memory hierarchy. We present three algorithms for inde...

متن کامل

Fast Cauchy-like and Singular Toeplitz-like Matrix Computations

computes the solution ~ x = T ?1 ~ b to a nonsingular Toeplitz or Toeplitz-like linear system T~ x = ~ b, a short displacement generator for the inverse T ?1 of T, and det T. We extend this algorithm to the similar computations with nn Cauchy (generalized Hilbert) and Cauchy-like matrices. Recursive triangular factorization of such a matrix can be computed by our algorithm at the cost of execut...

متن کامل

QR-factorization of displacement structured matrices using a rank structured matrix approach

A general scheme is proposed for computing the QR-factorization of certain displacement structured matrices, including Cauchy-like, Vandermonde-like, Toeplitz-like and Hankel-like matrices, hereby extending some earlier work for the QR-factorization of the Cauchy matrix. The algorithm employs a chasing scheme for the recursive construction of a diagonal plus semiseparable matrix of semiseparabi...

متن کامل

Cholesky Factorization of Semi-de nite Toeplitz Matrices

It can be shown directly from consideration of the Schur algorithm that any nn semi-deenite rank r Toeplitz matrix, T, has a factorization T = C r C T r with C r = C 11 C 12 0 0 where C 11 is r r and upper triangular. This paper explores the reliability of computing such a decomposition with O(nr) complexity using the Schur algorithm and truncating the Cholesky factor after computing the rst r ...

متن کامل

ذخیره در منابع من


  با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید

برای دانلود متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

ثبت نام

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

عنوان ژورنال:
  • Signal Processing

دوره 42  شماره 

صفحات  -

تاریخ انتشار 1995