Hindawi Publishing Corporation ξ e Scientiο¬c World Journal Volume 2014, Article ID 623294, 4 pages http://dx.doi.org/10.1155/2014/623294
Research Article The S-Transform of Distributions Sunil Kumar Singh Department of Mathematics, Rajiv Gandhi University, Doimukh, Arunachal Pradesh 791112, India Correspondence should be addressed to Sunil Kumar Singh; sks
[email protected] Received 30 August 2013; Accepted 10 October 2013; Published 2 January 2014 Academic Editors: B. Carpentieri and A. Ibeas Copyright Β© 2014 Sunil Kumar Singh. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. Parsevalβs formula and inversion formula for the S-transform are given. A relation between the S-transform and pseudodifferential operators is obtained. The S-transform is studied on the spaces S(Rπ ) and SσΈ (Rπ ).
1. Introduction The S-transform was first used by Stockwell et al. [1] in 1996. If π(π‘, π) is a window function, then the continuous S-transform of π(π‘) with respect to π is defined as [2] (Sπ π) (π, π) = β« π (π‘) π (π β π‘, π) πβπ2πβ¨π‘,πβ© ππ‘, Rπ
(1)
for π₯, π β Rπ . In signal analysis, at least in dimension π = 1, R2π is called the time-frequency plane, and in physics R2π is called the phase space. Equation (1) can be rewritten as a convolution as (Sπ π) (π, π) = (π (β
) πβπ2πβ¨β
,πβ© β π (β
, π)) (π) .
(2)
Now, we recall the definitions of Fourier transform on Rπ . Definition 1. If π(π‘) is defined on Rπ , then the Fourier transform of π is given by βπ2πβ¨π‘,πβ©
F [π (π‘)] (π) = β« π (π‘) π Rπ
where β¨π‘, πβ© =
βππ=1 π‘π ππ
ππ‘,
(3) π
is the usual inner product on R .
Definition 2. If π(π‘, π₯) is defined on R2π , then the partial Fourier transform of π(π‘, π₯) with respect to the first coordinate is given by F1 [π (π‘, π₯)] (π, π₯) = β« π (π‘, π₯) πβπ2πβ¨π‘,πβ© ππ‘, Rπ
(4)
and the partial Fourier transform of π(π‘, π₯) with respect to the second coordinate is given by F2 [π (π‘, π₯)] (π‘, π) = β« π (π‘, π₯) πβπ2πβ¨π₯,πβ© ππ₯. Rπ
(5)
Applying the convolution property for the Fourier transform in (2), we obtain (Sπ π) (π, π) = Fβ1 1 [F1 (π) (πΌ + π) F1 (π) (πΌ, π)] (π, π) , (6) where Fβ1 1 is the inverse Fourier transform. Now, we define the translation, modulation, and involution operators, respectively, by ππ π (π‘) = π (π‘ β π) ππ π (π‘) = ππ2πβ¨π,π‘β© π (π‘) Iπ (π‘) = π (βπ‘)
(translation) (modulation)
(7)
(involution) ,
π
where π‘, π, π β R . Definition 3 (the Dirac delta). The Dirac delta function is defined by β¬ π (π‘, π₯) πΏ (π‘, π₯) ππ‘ ππ₯ = π (0, 0) . Rπ
(8)
Definition 4 (tempered distribution). A function π β πΆβ (Rπ ) is said to be rapidly decreasing if σ΅¨ σ΅¨ πΎπΌ,π½ (π) = sup σ΅¨σ΅¨σ΅¨σ΅¨π₯πΌ π·π½ π (π₯)σ΅¨σ΅¨σ΅¨σ΅¨ < β, (9) π π₯βR
2
The Scientific World Journal
for all pairs of multi-indices πΌ, π½ β Nπ0 . The space of all rapidly decreasing functions on Rπ is denoted by S(Rπ ) or simply S. Elements in the dual space SσΈ of S are called tempered distribution.
2. Some Important Properties of S-Transform Some properties of S-transform can be found in [3β8] and certain properties of S-transform are obtained in this section. By definition, we have
This immediately implies the Plancherel formula σ΅© σ΅© σ΅© σ΅©σ΅© σ΅©σ΅©Sπ πσ΅©σ΅©σ΅©πΏ2 (Rπ ΓRπ ) = σ΅©σ΅©σ΅©πσ΅©σ΅©σ΅©πΏ2 (Rπ ) . Proof. Consider β¬
Rπ
(Sπ1 π1 ) (π, π) (Sπ2 π2 )(π, π) ππ ππ
=β¬
Rπ
(Sπ π) (π, π) = β« π (π β π‘, π) π (π‘) πβπ2πβ¨π‘,πβ© ππ‘
= β π1 (π‘) π2 (π₯) ππ2πβ¨(π₯βπ‘),πβ© Rπ
(10)
= β« π (βπ₯, π) πβπ π (π + π₯) ππ₯
Rπ
= β« π (βπ₯, π) πβπ πβπ π (π₯) ππ₯.
= β π1 (π‘) π2 (π₯) ππ2πβ¨(π₯βπ‘),πβ© πΏ (π₯ β π‘, 0) ππ ππ‘ ππ₯
Rπ
Rπ
Thus, the S-transform Sπ appears as a superposition of timefrequency shifts as follows:
Rπ
= β« π1 (π‘) π2 (π‘)ππ‘. Rπ
(11)
Example 5. If π(π‘, π) = π(π), that is, independent of π‘, then (Sπ π) (π, π) = β« π (π) π (π‘) πβπ2πβ¨π‘,πβ© ππ‘
(12)
= π (π) (Fπ) (π) .
Theorem 8 (inversion formula). If π β πΏ2 (Rπ ) and window function π satisfy the condition (14) of the previous theorem, then π (π‘) = β¬
Rπ
(Sπ π) (π, π) π (π β π‘, π)ππ2πβ¨π‘,πβ© ππ ππ.
So Sπ is a multiplication operator. In particular, if π(π‘, π) = 1, then (Sπ π)(π, π) = (Fπ)(π).
Proof. By the previous theorem we can write
Example 6. If π(π‘, π) = π(π‘), then
β¨π1 , π2 β© = β¬
(Sπ π) (π, π) = β« π (π β π‘) π (π‘) πβπ2πβ¨π‘,πβ© ππ‘
=β¬
Rπ
= β« πβπ π (βπ‘) π (π‘) πβπ2πβ¨π‘,πβ© ππ‘ Rπ
(Sπ π1 ) (π, π)
Rπ
Γ (β« π2 (π₯) π (π β π₯, π) πβπ2πβ¨π₯,πβ© ππ₯) ππ ππ Rπ
= β« (β¬
Rπ
Rπ
= F (ππβπ Iπ) (π) .
Rπ
Let π1 , π2 β πΏ2 (Rπ ) and let (Sπ1 π1 ) and (Sπ2 π2 ) be the Stransforms of π1 and π2 , respectively. Then β¬
Rπ
(Sπ1 π1 ) (π, π) (Sπ2 π2 ) (π, π)ππ ππ (15)
= β« π1 (π‘) π2 (π‘) ππ‘. Rπ
Rπ
(Sπ π1 ) (π, π) Γ π (π β π₯, π)ππ2πβ¨π₯,πβ© ππ ππ) π2 (π₯) ππ₯.
Theorem 7 (Parsevalβs formula). Let π1 and π2 be the window functions such that (14)
(18)
(Sπ π1 ) (π, π) (Sπ π2 ) (π, π) ππ ππ
Rπ
(13)
= β« πβπ Iπ (π‘) π (π‘) πβπ2πβ¨π‘,πβ© ππ‘
β« π1 (π β π‘, π) π2 (π β π₯, π) ππ = πΏ (π₯ β π‘, π β π) .
(17)
Γ (β« π1 (π β π‘, π) π2 (π β π₯, π)ππ) ππ ππ‘ ππ₯
Rπ
Rπ
Rπ
Rπ
= β« π (βπ₯, π) π (π + π₯) πβπ2πβ¨π+π₯,πβ© ππ₯
Sπ := β« π (βπ₯, π) πβπ πβπ ππ₯.
(β« π1 (π‘) π1 (π β π‘, π) πβπ2πβ¨π‘,πβ© ππ‘) Γ (β« π2 (π₯)π2 (π β π₯, π)πβπ2πβ¨π₯,πβ© ππ₯) ππ ππ
Rπ
Rπ
(16)
(19) Hence π1 (π‘) = β¬
Rπ
(Sπ π1 ) (π, π) π (π β π‘, π)ππ2πβ¨π‘,πβ© ππ ππ.
(20)
Definition 9. Let π be a window function and Sπ is the Stransform. Then the transform Sβπ defined by β¨Sπ π, πβ© = β¨π, Sβπ πβ©
(21)
The Scientific World Journal
3
is called the adjoint of Sπ . If π β πΏ2 (Rπ ) and π β πΏ2 (Rπ Γ Rπ ), then (21) implies that (Sβπ π) (π‘) = β¬ π (π, π) π (π β π‘, π)ππ2πβ¨π‘,πβ© ππ ππ, Rπ
(22)
where π‘, π, π β Rπ . Theorem 10 (Parsevalβs formula for Sβπ ). Let π1 and π2 be the window functions that satisfy the condition (14). If π1 , π2 β πΏ2 (Rπ Γ Rπ ), then
(23) = β¬ π1 (π, π) π2 (π, π) ππ ππ, Rπ
and the Plancherel formula is σ΅© σ΅© σ΅©σ΅© β σ΅©σ΅© σ΅©σ΅©Sπ πσ΅©σ΅©πΏ2 (Rπ ) = σ΅©σ΅©σ΅©πσ΅©σ΅©σ΅©πΏ2 (Rπ ΓRπ ) .
(24)
Proof. Consider β« (Sβπ1 π1 ) (π‘) (Sβπ2 π2 ) (π‘) ππ‘ Rπ
= β« (β¬ π1 (π, π) π1 (π β π‘, π)ππ2πβ¨π‘,πβ© ππ ππ) Rπ
Rπ
Γ (β¬ π2 (πΌ, π) π2 (πΌ β Rπ
Sβπ = Sβ1 π . This proves the theorem.
Definition 12 (pseudodifferential operator). Let π be a (measurable) function or a tempered distribution on Rπ . Then the operator πΎπ π (π‘) = β« π (π‘, π) πΜ (π) ππ2πβ¨π‘,πβ© ππ Rπ
(29)
The pseudodifferential operator plays an important role in the theory of partial differential equations. The pseudodifferential operator has been studied on function and distribution spaces by many authors. Details of the concept can be found in [9, 10]. 2.1. Relation between the S-Transform and Pseudodifferential Operator. Here we give a direct relation between S-transform and pseudodifferential operator which will may be very useful in the study of S-transform of distribution spaces. The continuous S-transform of a function π with respect to a window function π is given by (Sπ π) (π, π) = β« π (π‘) π (π β π‘, π) πβπ2πβ¨π‘,πβ© ππ‘ Rπ
π‘, π)ππ2πβ¨π‘,πβ© ππΌ ππ) ππ‘
= β« π (π β π₯) π (π₯, π) πβπ2πβ¨(πβπ₯),πβ© ππ₯ Rπ
= β«β π1 (π, π) π2 (πΌ, π)ππ2πβ¨π‘,πβπβ©
= πβπ2πβ¨π,πβ© β« πβπ π (βπ₯) π (π₯, π) ππ2πβ¨π₯,πβ© ππ₯
Rπ
Γ (β« π1 (π β π‘, π)π2 (πΌ β π‘, π) ππ‘) ππ ππ ππΌ ππ Rπ
Rπ
= πβπ2πβ¨π,πβ© β« F [F (πβπ π)] (π₯) Rπ
= β«β π1 (π, π) π2 (πΌ, π)ππ2πβ¨π‘,πβπβ©
Γ π (π₯, π) ππ2πβ¨π₯,πβ© ππ₯
Rπ
Γ πΏ (π β πΌ, π β π) ππ ππ ππΌ ππ
= πβπ2πβ¨π,πβ© πΎπ [F (πβπ π)] (π) , (30)
= β¬ π1 (π, π) π2 (π, π) ππ ππ. Rπ
(28)
is called the pseudodifferential operator.
β« (Sβπ1 π1 ) (π‘) (Sβπ2 π2 ) (π‘) ππ‘ Rπ
Thus
(25)
where π(π, π₯) = π(π₯, π).
This proves the theorem.
3. The S-Transform of Distributions
Theorem 11. If the window function π satisfies the condition (14), then
In this section we will investigate the S-transform of tempered distribution by means of the Fourier transform.
Sπ Sβπ = πΌ = Sβπ Sπ ,
Theorem 13. If π β S(R2π ), then Sπ maps S(Rπ ) into S(R2π ).
(26)
where πΌ is the identity operator.
Proof. By (6) we have (Sπ π) (π, π) = Fβ1 1 [F1 (π) (πΌ + π) F1 (π) (πΌ, π)] (π, π) . (31)
Proof. By definition Sβπ [Sπ1 π] (π‘) = β¬
Rπ
(Sπ1 π) (π, π) π (π β π‘, π)eπ2πβ¨π‘,πβ© ππ ππ
= Sβ1 π [Sπ1 π] (π‘) . (27)
Thus, (Sπ π) β S(R2π ), since the Fourier transform is continuous isomorphism from S(Rπ ) to S(Rπ ), and its inverse is also a continuous isomorphism from S(Rπ ) to S(Rπ ) (see [11], page 66-67).
4
The Scientific World Journal
Theorem 14. If π β S(R2π ), then Sπ maps SσΈ (Rπ ) into SσΈ (R2π ). Proof. For any π β SσΈ (Rπ ) and π β S(R2π ), we have β¨(Sπ π) (π, π) , π (π, π)β© =β¬
Rπ
(β« π (π‘) π (π β π‘, π) πβπ2πβ¨π‘,πβ© ππ‘) π(π, π) ππ ππ Rπ
(in fact π = π, since π β S) = β« π (π‘) (β¬ π(π, π)π (π β π‘, π) πβπ2πβ¨π‘,πβ© ππ ππ) ππ‘ Rπ
Rπ
= β¨π, πβ© , (32) where π (π‘) = β¬ π (π, π) π (π β π‘, π)ππ2πβ¨π‘,πβ© ππ ππ Rπ
= β« (π β (Iπ)) (π‘, π) ππ2πβ¨π‘,πβ© ππ
(33)
Rπ
π = Fβ1 2 (π β (Iπ)) (π‘) β S (R ) .
Thus, (Sπ π)(π, π) β SσΈ (R2π ). Theorem 15. If π β SσΈ (R2π ), then Sπ maps S(Rπ ) into SσΈ (R2π ). Proof. If π β S(Rπ ) and π β S(R2π ), then ππ,π (π, π) := F (π) (πΌ + π)π (πΌ, π) β S (R2π ) .
(34)
Thus for any π β SσΈ (R2π ), we have β¨(F1 π) (πΌ, π) , ππ,π (πΌ, π)β© =β¬ =β¬
Rπ
Rπ
(F1 π) (πΌ, π) (Fπ) (πΌ + π) π (πΌ, π) ππΌ ππ (35) (F1 (Sπ π)) (πΌ, π) π (πΌ, π) ππΌ ππ
= β¨(F1 (Sπ π)) (πΌ, π) , π (πΌ, π)β© . Thus F1 (Sπ π) β SσΈ (R2π ) and hence (Sπ π) β SσΈ (R2π ).
Conflict of Interests The authors declare that there is no conflict of interests.
Acknowledgment The author expresses his sincere thanks to Professor R. S. Pathak for his help and encouragement.
References [1] R. G. Stockwell, L. Mansinha, and R. P. Lowe, βLocalization of the complex spectrum: the S transform,β IEEE Transactions on Signal Processing, vol. 44, no. 4, pp. 998β1001, 1996.
[2] S. Ventosa, C. Simon, M. Schimmel, J. J. Danobeitia, and A. Manuel, βThe S-transform from a wavelet point of view,β IEEE Transactions on Signal Processing, vol. 56, no. 7, pp. 2771β2780, 2008. [3] S. K. Singh, βThe S-Transform on spaces of type S,β Integral Transforms and Special Functions, vol. 23, no. 7, pp. 481β494, 2012. [4] S. K. Singh, βThe S-Transform on spaces of type W,β Integral Transforms and Special Functions, vol. 23, no. 12, pp. 891β899, 2012. [5] S. K. Singh, βThe fractional S-Transform of tempered ultradistibutions,β Investigations in Mathematical Sciences, vol. 2, no. 2, pp. 315β325, 2012. [6] S. K. Singh, βThe fractional S-Transform on spaces of type S,β Journal of Mathematics, vol. 2013, Article ID 105848, 9 pages, 2013. [7] S. K. Singh, βThe fractional S-Transform on spaces of type W,β Journal of Pseudo-Differential Operators and Applications, vol. 4, no. 2, pp. 251β265, 2013. [8] S. K. Singh, βA new integral transform: theory part,β Investigations in Mathematical Sciences, vol. 3, no. 1, pp. 135β139, 2013. [9] K. Grochenig, Foundations of Time-Frequency Analysis, BirkhΒ¨aauser, Boston, Mass, USA, 2001. [10] S. Zaidman, Distributions and Pseudo-Differential Operators, Logman, Essex, UK, 1991. [11] R. S. Pathak, A Course in Distribution Theory and Applications, Narosa Publishing House, New Delhi, India, 2009.