Skip to main content

Unified Theory: Frequency Domain Analysis

  • Chapter
Unified Signal Theory
  • 1595 Accesses

Abstract

In this chapter, signals are studied in the frequency domain by means of the Fourier transform, which is introduced in unified form using the Haar integral as

$$S(f) = \int_I \mathrm{d} t\, s(t) \psi^{\ast}(f,t), \quad f\in \widehat{I},$$

where I is the signal domain, \(\widehat{I}\) is the frequency domain and ψ (f,t) is the conjugate of the kernel. A preliminary problem is the identification of the frequency domain \(\widehat{I}\) and of the kernel ψ(f,t). In particular, for the groups of ℝ, we shall find that the kernel has the familiar exponential form ψ(f,t)=exp (i2πft). Since the frequency domain \(\widehat{I}\) associated to a given LCA group I is still an LCA group (the dual group), all definitions and operations introduced in the signal domain (in particular the Haar integral) are straightforwardly transferred to the frequency domain. For the Fourier transform, several rules will be established in a unified form and then applied to specific classes of signals, one-dimensional as well as multidimensional.

The chapter also contains two unusual topics: the Fourier transform on multiplicative groups and the fractional Fourier transform.

This is a preview of subscription content, log in via an institution to check access.

Access this chapter

Chapter
USD 29.95
Price excludes VAT (USA)
  • Available as PDF
  • Read on any device
  • Instant download
  • Own it forever
eBook
USD 84.99
Price excludes VAT (USA)
  • Available as EPUB and PDF
  • Read on any device
  • Instant download
  • Own it forever
Softcover Book
USD 109.99
Price excludes VAT (USA)
  • Compact, lightweight edition
  • Dispatched in 3 to 5 business days
  • Free shipping worldwide - see info
Hardcover Book
USD 109.99
Price excludes VAT (USA)
  • Durable hardcover edition
  • Dispatched in 3 to 5 business days
  • Free shipping worldwide - see info

Tax calculation will be finalised at checkout

Purchases are for personal use only

Institutional subscriptions

Notes

  1. 1.

    This can be done without ambiguity in most of the cases. However, in some algebraic steps involving matrices, the m-tuples t=(t 1,…,t m ) and f=(f 1,…,f m ) must be interpreted as column vectors, and ft must be replaced by ft=f 1 t 1+⋅⋅⋅+f m t m , where f′ is the transpose of f.

  2. 2.

    Simultaneous finite duration and finite bandwidth may be incompatible (see Sect. 9.5), so that one of the definitions must be relaxed using conventional duration and bandwidth (see Sect. 13.11).

  3. 3.

    The technique can be used to define other fractional operators. In [7], it was used to define the fractional DCT.

  4. 4.

    The dual group \(\widehat{I}\) is not unique, but two different dual groups are isomorphic.

References

  1. M. Abramowitz, I.A. Stegun, Handbook of Mathematical Functions (Dover, New York, 1970)

    Google Scholar 

  2. L.B. Almeida, The fractional Fourier transform and time–frequency representations. IEEE Trans. Signal Process. 42, 3084–3091 (1994)

    Article  Google Scholar 

  3. H. Bremermann, Distribution, Complex Variables, and Fourier Transforms (Addison–Wesley, Reading, 1965)

    Google Scholar 

  4. C. Candan, M.A. Kutay, H.M. Ozaktas, The discrete fractional Fourier transform. IEEE Trans. Signal Process. 48, 1329–1337 (2000)

    Article  MathSciNet  MATH  Google Scholar 

  5. G. Cariolaro, T. Erseghe, P. Kraniauskas, N. Laurenti, A unified framework for the fractional Fourier transform. IEEE Trans. Signal Process. 46, 3206–3219 (1998)

    Article  MathSciNet  MATH  Google Scholar 

  6. G. Cariolaro, T. Erseghe, P. Kraniauskas, N. Laurenti, Multiplicity of fractional Fourier transforms and their relationships. IEEE Trans. Signal Process. 48, 227–241 (2000)

    Article  MathSciNet  MATH  Google Scholar 

  7. G. Cariolaro, T. Erseghe, The fractional discrete cosine transform. IEEE Trans. Signal Process. 50, 902–911 (2002)

    Article  MathSciNet  Google Scholar 

  8. T. Erseghe, G. Cariolaro, An orthonormal class of exact and simple DFT eigenfunctions with a high degree of symmetry. IEEE Trans. Signal Process. 51, 2527–2539 (2003)

    Article  MathSciNet  Google Scholar 

  9. I.B. Gel’fand, G.E. Shilov, Generalized Functions. Applications of Harmonic Analysis, vol. 4 (Academic Press, Dordrecht, 1964)

    MATH  Google Scholar 

  10. E. Hewitt, K.A. Ross, Abstract Harmonic Analysis, vols. 1, 2 (Springer, Berlin, 1963)

    Book  Google Scholar 

  11. M.J. Lighthill, Introduction to Fourier Analysis and Generalized Functions (Cambridge University Press, Cambridge, 1958)

    Book  Google Scholar 

  12. A.W. Lohmann, Image rotation, Wigner rotation, and the fractional Fourier transform. J. Opt. Soc. Am. A 10(10), 2181–2186 (1993)

    Article  MathSciNet  Google Scholar 

  13. A.W. Lohmann, B.H. Soffer, Relationship between the Radon–Wigner and fractional Fourier transforms. J. Opt. Soc. Am. A 11(6), 1798–1801 (1994)

    Article  MathSciNet  Google Scholar 

  14. L.H. Loomis, An Introduction to Abstract Harmonic Analysis (Van Nostrand, New York, 1953)

    MATH  Google Scholar 

  15. D. Mendlovic, H.M. Ozaktas, Fractional Fourier transforms and their optical implementation: I. J. Opt. Soc. Am. A 10(9), 1875–1881 (1993)

    Article  Google Scholar 

  16. V. Namias, The fractional order Fourier transform and its applications to quantum mechanics. Inst. Math. Appl. 25, 241–265 (1980)

    Article  MathSciNet  MATH  Google Scholar 

  17. H.M. Ozaktas, D. Mendlovic, Fractional Fourier transforms and their optical implementation: II. J. Opt. Soc. Am. A 10(12), 2522–2531 (1993)

    Article  MathSciNet  Google Scholar 

  18. H.M. Ozaktas, B. Barshan, D. Mendlovic, L. Onural, Convolution, filtering and multiplexing in fractional Fourier domains and their relation to chirp and wavelet transforms. J. Opt. Soc. Am. A 11(2), 547–559 (1994)

    Article  MathSciNet  Google Scholar 

  19. A. Papoulis, Systems and Transforms with Applications in Optics (McGraw–Hill, New York, 1968)

    Google Scholar 

  20. S.C. Pei, M.H. Yeh, C.C. Tseng, Discrete fractional Fourier transform based on orthogonal projections. IEEE Trans. Signal Process. 47, 1335–1348 (1999)

    Article  MathSciNet  MATH  Google Scholar 

  21. L. Pontryagin, Topological Groups (Princeton University Press, Princeton, 1946)

    Google Scholar 

  22. S. Roman, Advance Linear Algebra (Springer, New York, 1992)

    Book  Google Scholar 

  23. W. Rudin, Fourier Analysis on Groups (Interscience, New York, 1962)

    MATH  Google Scholar 

  24. E.C. Titchmars, Introduction to the Theory of Fourier Integrals (Oxford University Press, New York, 1937)

    Google Scholar 

Download references

Author information

Authors and Affiliations

Authors

Corresponding author

Correspondence to Gianfranco Cariolaro .

Appendices

Appendix A: Fourier Kernel from Characters

A character of an LCA group I is every continuous complex function on I, ψ f (t), tI, with the properties (see (5.4) and (5.5))

$$ \psi _f \left( {t_1+ t_2 } \right) = \psi _f \left( {t_1 } \right)\psi _f \left( {t_2 } \right),\,\,t_1 ,t_2\in I, $$
(5.130a)
$$ \left| {\psi _f \left( t \right)} \right| = 1,\,t \in I. $$
(5.130b)

The index f identifies a specific character ψ f ; the collection of all the indexes f identifies a set, which represents the dual group \(\widehat{I}\) (the dual of I).Footnote 4 In fact, it can be shown that the index set is always an LCA group [23]. Now, from the set of characters ψ f (t), with tI and \(f\in \widehat{I}\), we can define a function of two variables ψ(f,t)=ψ f (t), which represents the kernel of the FT on I.

From conditions (5.130a), (5.130b) the following important properties for the kernel ψ(f,t), which generalize the properties of the exponential ei2πft, can be established

$$ \psi \left( {f,t_1+ t_2 } \right) = \psi \left( {f,t_1 } \right)\psi \left( {f,t_2 } \right), $$
(5.131a)
$$ \psi \left( {f,0} \right) = 1,\,\,\psi \left( {0,t} \right) = 1, $$
(5.131b)
$$ \psi \left( {f,t} \right) = \,\psi *\left( {f, - t} \right) = 1/\psi \left( {f, - t,} \right) $$
(5.131c)
$$ \psi \left( {f, - t} \right) = \psi \left( { - f,t} \right), $$
(5.131d)
$$ \psi \left( {f_1+ f_2 ,t} \right) = \psi \left( {f_1 ,t} \right)\psi \left( {f_2 ,t} \right). $$
(5.131e)

For instance, the first of (5.131b) is proved by fixing in (5.130a) t 1=t 2=0; and (5.131c) is proved by letting t 1=−t 2=t. We see that the separability with respect to t leads to the separability with respect to f. Note that in (5.130a), (5.130b) and (5.131a), (5.131b), (5.131c), (5.131d), (5.131e) the group operation + is not necessarily the same for I and \(\widehat{I}\). In (5.131a), + is the operation on I, whereas in (5.131e) + is the operation on \(\widehat{I}\). Finally, in (5.131b) the first 0 is the identity element of I and the second 0 is the identity element of \(\widehat{I}\).

Fourier Kernel on ℝ

When I=ℝ in (5.130a), the group operation + is the ordinary addition and the only continuous solution to the functional equation ψ(t 1+t 2)=ψ(t 1)ψ(t 2) has the form ψ(t)=ept with p an arbitrary complex number. But, constraint (5.130b) restricts p to be imaginary and this is sufficient to conclude that \(\{\mathrm{e}^{\mathrm{i}at} \:|\: a\in \mathbb{R}\}\) is the character set and that ℝ is the dual of ℝ. On the other hand, we can denote the real number a in the form a=2πf with f still real, thus obtaining the form (5.6), that is, ψ (f,t)=ei2πft.

Fourier Kernel on ℝm

If a group is separable, I=I 1×I 2, the Fourier kernel on I is given by the product of the kernels on I 1 and I 2 (see [23]). Since ℝ2 is separable, from this rule we find that the kernel on ℝ2 is given by

$$\psi_{\mathbb{R}^2}(f_1,f_2;t_1,t_2)=\psi_{\mathbb{R}}(f_1,t_1) \psi_{\mathbb{R}}(f_2,t_2)=\mathrm{e}^{\mathrm{i}2\pi (f_1 t_1+f_2 t_2)}.$$

The same form holds for all the subgroups of ℝ2, also not separable, since they have the group operation in common with ℝ2. The generalization to ℝm is immediate and gives (5.4).

Appendix B: Invertibility of the Fourier Transform

We prove that, from the FT calculated using (5.1), we can recover the signal using (5.2). Denoting the inverse transform by

$$\tilde{s}(t) = \int_{\widehat{I}} \mathrm {d}f\; S(f) \psi^{\ast}(f,t) ,$$

and substituting the expression of S(f) given by (5.1), we get

$$\tilde{s}(t) = \int_{\widehat{I}} \mathrm {d}f\; \int_I \mathrm {d}u\; s(u)\psi^*(f,u) \psi(f,t),$$

where, from properties (5.131a), (5.131b), (5.131c), (5.131d), (5.131e) ψ (f,u)ψ(f,t)=ψ(f,−u)ψ(f,t)=ψ(f,tu). Hence

$$\tilde{s}(t) = \int_I \mathrm {d}u\; s(u) \int_{\widehat{I}} \mathrm {d}f \;\psi(f,t-u)= \int_{I} \mathrm {d}u \, s(u) \delta_I(t-u) = s(t) ,$$

where we used the orthogonality condition (5.10a), (5.10b) and the sifting property of the impulse (4.76).

Appendix C: Proof of Theorem 5.1 on the Dual Group

We prove that if the function ψ f (t)=ψ(f,t) (with fixed f) is defined on I 0 and has periodicity P, then the function ψ t (f)=ψ(f,t) (with fixed t) is defined on P and has periodicity \(I_{0}^{\star }\). The periodicity of ψ f (t) yields ψ f (t+u)=ψ f (t), uP, and, recalling that ψ f (t+u)=ψ f (t)ψ f (u), it follows that

$$ \psi_f(u) = 1 , \quad u \in P .$$
(5.132)

Thus, any frequency f satisfying (5.132) ensures the periodicity P with respect to the time domain u. The set of such frequencies is \(P^{\star}=\{f \:|\: \psi_{f}(u) = 1, \ u\in P\}\), which defines the domain of the kernel with respect to the frequency f.

Analogously, the possible periodicity of the function ψ t (f) with respect to f is expressed by the condition ψ t (f+v)=ψ t (f), tI 0, which is equivalent to ψ t (v)=1, tI 0. The frequencies v that verify this condition form the reciprocal of I 0. Therefore, the periodicity of the dual group is given by \(I_{0}^{\star }\).

Appendix D: Proof of Theorem 5.2 on the Representation of the Dual Group

To find the representation of the reciprocal J of an mD group J, we note that the condition tJ in (5.31) can be expressed in the form t=Jh, hH. Hence, we have

$$J^\star = \{\textbf{f} \mid \mathbf{f}'\mathbf{Jh}\in \mathbb{Z}, \ \textbf{h}\in H\}.$$

Now, if we compare the expression of J with the expression of H , given by

$$H^\star = \{\textbf{v} \mid \textbf{vh}\in \mathbb{Z},\ \textbf{h}\in H\},$$

we find the relation between the frequencies v of H and the frequencies f of J , namely

$$\textbf{v}'=\textbf{f}' \textbf{J} \rightarrow \textbf{f} = (\textbf{J}')^{-1} \textbf{v} , \quad \textbf{v} \in H^{\star }.$$

This completes the proof.

Appendix E: Proof of Poisson’s Summation Formula

For the proof we use the theory of elementary transformations and particularly the Duality Theorem of Sect. 6.13. Let us consider a signal s(t), tI, that is, IU down-sampled, where I=I 0/PU=U 0/P and U 0 is a lattice. The relation is

$$y(t) = s(t) , \quad t \in U $$
(5.133a)

where y(t) is the down-sampled signal (Fig. 5.28).

Fig. 5.28
figure 28

Diagram for the proof of Poisson’s summation formula

In the frequency domain, we have an \(\widehat{I} \rightarrow \widehat{U}\) up-periodization, where \(\widehat{I}=P^{\star }/I_{0}^{\star }\), \(\widehat{U}= P^{\star }/U_{0}^{\star }\). The relation is (see (6.80))

$$ Y(f) = \sum_{\lambda\in U_0^\star /I_0^\star } S(f-\lambda) =\sum_{\lambda\in U_0^\star /I_0^\star } S(f+\lambda),\quad f\in \widehat{I}.$$
(5.133b)

Finally, the application of the rule area (y)=Y(0) gives

$$ \sum_{u\in U_0/P} \mathrm {d}(U_0)s(u) =\sum_{\lambda\in U_0^\star /I_0^\star } S(\lambda) .$$
(5.134)

This complete the proof.

Rights and permissions

Reprints and permissions

Copyright information

© 2011 Springer-Verlag London Limited

About this chapter

Cite this chapter

Cariolaro, G. (2011). Unified Theory: Frequency Domain Analysis. In: Unified Signal Theory. Springer, London. https://doi.org/10.1007/978-0-85729-464-7_5

Download citation

  • DOI: https://doi.org/10.1007/978-0-85729-464-7_5

  • Publisher Name: Springer, London

  • Print ISBN: 978-0-85729-463-0

  • Online ISBN: 978-0-85729-464-7

  • eBook Packages: EngineeringEngineering (R0)

Publish with us

Policies and ethics