Fractional order signal processing. Introductory concepts and applications. (Q640555)

This short monograph addresses several techniques involving signals, processes, and operations that are of {\textit{fractional order}}. While L'Hospital first asked Leibniz about the possibility of fractional differentiation in 1695 and applications to integral equations already occurred in 1823 (see e.g. [\textit{B. Ross}, ``The development of fractional calculus 1695--1900'', Hist. Math. 4, 75--89 (1977; Zbl 0358.01008)]), physical interpretations are more recent, e.g. [\textit{I. Podlubny}, ``Geometric and physical interpretation of fractional integration and fractional differentiation'', Fract. Calc. Appl. Anal. 5, No. 4, 367--386 (2002; Zbl 1042.26003)]. The monograph discussed here should be particularly appealing to a reader who wants a brief introduction to matters that involve long-range dependence that arises in a number of processes such as network traffic, DNA sequences, stock prices etc., and a rudimentary understanding of which methods are appropriate for identifying and characterizing such behavior. Chapter one defines fractional derivatives and integrals and briefly outlines their geometric and physical interpretations, and their uses in science and engineering. The second chapter addresses fractional order signals and systems, including filtering issues such as how to express fractional order transfer functions in terms of Laplace transforms, numerical implementations, and discrete-time fractional order models. Methods for approximating transfer functions include Carlson's method, a recursive scheme that applies to a transfer function that can be expressed as a rational fractional power of a rational function, Charif's method for iteratively approximating \((1+s/p)^m\) with pole \(p\) and fractional power \(m\), and Oustaloup's method which provides a good fit to fractional order elements in a given frequency band. Discrete-time fractional order models discussed include auto-regressive fractional integrated moving average (ARFIMA) and fractional autoregressive integrated moving average (FARIMA) models that generalize standard ARMA models. Chapter 3 addresses fractional order processes, including fractional Brownian motion and fractional and multi-fractional Gaussian noise, and the notion of fractional dimension and Hurst exponent and time-varying versions. The discussion does not extend to the fractal spectrum of general multi-fractional processes. Statistical models are discussed as well. Chapters 5 and 6 provide more in-depth treatment of the concepts introduced in Chapters 2 and 3, respectively. Chapter 5 deals with fractional order system identification, including time-domain and frequency-domain methods and output error models. Specific methods include Levy's identification method for fractional order systems which minimizes the squared norm of the error between a system frequency response \(G\) and ratios \(N/D\) of polynomials in the variable \(s^q\) where \(q\) is a given fractional power. Variants due to Vinagre and Sanathanan and Koerner are also discussed. Chapter 6 introduces fractional Kalman filtering methods and also discusses fractional low-order moments and their applications. The major issues that arise in numerical implementations are mentioned briefly, but in-depth analysis of numerical errors such as can be found in the referenced literature is not repeated here. One slightly puzzling aspect is how the Fractional Fourier Transform (FrFT), the subject of Chapter 4, along with other fractional transformations, and fractional convolution, introduced in Chapter 2 via the FrFT, fit in with the other fractional order methods. That is, the term {\textit{fractional}} has a different meaning in regards to Fourier analysis, in which it corresponds to a rotation of the time-frequency plane, than it does in regard to differentiation, in which it can be defined as division by a fractional power of the Fourier variable with respect to the standard Fourier transform.