Journal of Signal and Information Processing, 2013, 4, 120-124 doi:10.4236/jsip.2013.43B021 Published Online August 2013 (http://www.scirp.org/journal/jsip) Discrete Entropic Uncertainty Relations Associated with FRFT* Guanlei Xu1, Xiaotong Wang2, Lijia Zhou1, Limin Shao1, Xiaogang Xu2 1Ocean department of Dalian Naval Academy, Dalian, China, 116018; 2Navgation department of Dalian Naval Academy, Dalian, China, 116018. Email: xgl_86@163.com Received March, 2013. ABSTRACT Based on the definition and properties of discrete fractional Fourier transform (DFRFT), we introduced the discrete Hausdorff-Young inequality. Furthermore, the discrete Shannon entropic uncertainty relation and discrete Rényi en- tropic uncertainty relation were explored. Also, the condition of equality via Lagrange optimization was developed, as shows that if the two conjugate variables have constant amplitudes that are the inverse of the square root of numbers of non-zero elements, then the uncertainty relations reach their lowest bounds. In addition, the resolution analysis via the uncertainty is discussed as well. Keywords: Discrete Fractional Fourier Transform (DFRFT); Uncertainty Principle; Rényi Entropy; Shannon En tropy 1. Introduction Uncertainty principle not only holds in analog signals, but also in discrete signals [1,2]. Recently, with the de- velopment of fractional Fourier transform (FRFT), ana- log generalized uncertainty relations associated with FRFT have been carefully explored in some papers such as [3,4,16], which effectively enrich the ensemble of FRFT. However, up till now there has been no reported article covering the d iscrete generalized uncertainty rela- tions associated with discrete fractional Fourier trans- form (DFRFT). From the viewpoint of engineering ap- plication, discrete data are widely used and seem to be more profitable than the analog ones. Hence, there is enough need to explore discrete generalized uncertainty relations. DFRFT is the discrete version of FRFT [5,6], which is applied in practical engineering fields. In this article we will discuss the entropic uncertainty relations [7,8] associated with DFRFT. In this article, we made some contributions such as follows. The first contribu- tion is that we extend the traditional Hausdorff-Young inequality to the FRFT domain with finite supports. It is shown that these bounds are connected with lengths of the supports and FRFT parameters. The second contribu- tion is that we derived the Shannon entropic uncertainty principle in FRFT domain for discrete case, based on which we also derived the conditions when these uncer- tainty relations have the equalities via Lagrange optimi- zation. The third contribution is that we derived the Renyi entropic uncertain ty principle in FRFT domain for discrete case. The final contribution is that we discussed the resolution in multiple FRFT domains as a succession of above derivative, including new proofs. In a word, there have been no reported papers covering these gener- alized discrete entropic uncertainty relations on FRFT. 2. Preliminaries 2.1. FRFT and DFRFT First, confirm that you have the correct template for your paper size. This template has been tailored for output on the custom paper size (21 cm * 28.5 cm). Before discussing the uncertainty principle, we intro- duce some relevant preliminaries. Here we first briefly review the definition of FRFT. For given analog signal and )()()( 21 RLRLtf 1)( 2tf , its FRFT [5,6] is defined as ()()()(,) uFftftK utdt 22 cot cot sin 22 (1cot) /2() () 2 ()(2 1) iut iu it ieeeftdt ft n ft n n , (1) where Z nand is the complex unit, i is the trans- form parameter defined as that in [5,6]. In addition, *This work was fully supported by the NSFC (61002052) and partly supported by the N SFC (61250006). Copyright © 2013 SciRes. JSIP
Discrete Entropic Uncertainty Relations Associated with FRFT 121 . If , )()( tftfFF dutuKu ),()( N CNx )(,),3 , )()( tfFtfFF i.e., the inverse FRFT reads: . Ftf )( xxx ),2(),1( Let N be a discrete time series with cardinality N and (xxxxX ,,,, 321 1 2. Assume its DFRFT under the transform parameter X N NCx ˆ , xxx , ˆ , ˆ , ˆ321 X ˆ . Then the DFRFT [5] can be written as 2 2co 2 )/ ik 2 cot t sin 2 cot () in ikn NN 1 ˆ() (1 N l ki 1(,) ( N luknx N kn eeexn , )nN ,1 XU . (2) Also, we can rewrite the definition (2) as X ˆ, where . NN nkuU ),( Clearly, for DFRFT we have the following proper- ties[5]: XUX k XU UUXUU UU 2, XX UX k 2 , 1 2X ˆ 2UX . In the follows, we will assume that the transform pa- rameter 20 and . The max difference between the discrete and analog definitions is the support: one is finite and discrete and the other one is infinite and analog. 2.2. Shannon Entropy and Rényi Entropy For any discrete random variable () and its probability density function , the Shannon en- tropy [9] and the Rényi Entropy [10] is defined as, re- spectively n x) n xNn ,,1 (p )(ln)( 1n N nnn xpxpxH , N nnn xpxH 1)(ln 1 1 . Hence, in this article, we know that for any DFRFT , the Shannon entropy and the Rényi Entropy associated with DFRFT is defined as, respectively N NCxxxxX ˆ ,, ˆ , ˆ , ˆ ˆ321 2 2 1)( ˆ ln)( ˆˆnxnxxH N n , N nnxxH 1 2 )( ˆ ln 1 1 ˆ . Clearly, if ,1 xHxH ˆˆ . 2.3. Discrete Hausdorff-Young Inequality Associated with DFRFT Let be a discrete time series with cardinality N and its DFRFT with and the transform parameter N NCNxxxxxxxxX )(,),3(),2(),1(,,,, 321 N NCxxxxX ˆ ,, ˆ , ˆ , ˆ ˆ321 XUX ˆ . Since 1 2X, 1 ˆ2 2 XUX from Parseval’s theorem. Here 2 1 2 1 2 N nn xX . Clearly, we can ob- tain 1 XMXU UM with . )(sup luU r l Here with NlluUr,,1,)( . Hence, we have 1 XMXU with UM. Then from Riesz’s theorem [11,12], we can obtain the discrete Hausdorff-Young inequality p p p qXMXU 2 with and 21 p1 11 qp . Set UU , then , we obtain UUUU p p p qXMXUU 2 with UUUM . Let XUY , then . In addition, from the property of DFRFT we can have YUX )sin( 1 N UUM. Hence we obtain p p p qYUMYU 2 with )sin( 1 N M. Since the value of can be taken arbitrarily in , can also be taken arbitrarily in . Therefore, we can obtain the following lemma. N C YN C Lemma 1: For any given discrete time series N NCNxxxxxxxxX )(,),3(),2(),1(,,,, 321 with cardinality N and 1 2, ( ) is the DFRFT transform matrix associated with the transform parameter XU U ( , respectively), then we can obtain the generalized discrete Hausdorff-Young inequality p p p qXUNXU 2 2 )sin( with 21 p and 1 11 qp. Clearly, this is the discrete version of Hausdorff- Young inequality. In the next sections, we will use this lemma to prove the new uncertainty relations. 3. The Uncertainty Relations 3.1. Shannon Entropic Principle Theorem 1: For any given discrete time series )(,),3(),2(),1(,,,, 321NxxxxxxxxX N CN with cardinality N and 1 2, ( ) is the DFRFT series associated with the transform parameter Xx ˆx ˆ ( , respectively), ( N) counts the non-zero elements of ( , respectively), then we can obtain the gener- alized discrete Shannon entropic uncertainty relation N x ˆx ˆ )sin(ln ˆˆ NxHxH , (3) Copyright © 2013 SciRes. JSIP
Discrete Entropic Uncertainty Relations Associated with FRFT 122 where N nnxnxxH 1 22 )( ˆ )( ˆ ln ˆ , N mmxmxxH 1 22 )( ˆ )( ˆ ln ˆ , which are Shannon entropies. The equality in (3) holds iff N x1 ˆ and N x1 ˆ. Proof: From lemma 1, we have 1 )( ˆ )( ˆ )sin( 1 11 1 1 2 2 p p N np p p N m p p p nx mxN . Take natural logarithm in both sides in above inequal- ity, we can obtain 0)( pT , where 1 21 ˆ ()lnsin()ln() 2 p N m p TpNx m pp 1 1 1ˆ ln() p Np n pxn p . Since and 21 p1 2X,we know 0)2( T0)( . Note if . Hence, 0)( pT 21 p' pT if . Since 2p 22 1 1 1 1 21 1 1 1 1 11 ˆ '( )lnsin()ln() ˆˆ ln ()() 1 ˆ() 1ˆ ln() ˆˆ ()ln () 1, (1) ˆ() N m p N m p N m p Np n p Np n p Np n Tp Nxm pp xm xm pxm xn p xn xn pp xn we can obtain the final result in theorem 1. Now consider when the equality holds. From theorem 1, that the equality holds in (3) implies that xHxH ˆˆ reaches its minimum bound, which means that Minimize xHxH ˆˆ subject to 1 ˆˆ 2 2 xx, i.e. Minimize N m N nmxmxnxnx 1 22 1 22 )( ˆ )( ˆ ln)( ˆ )( ˆ ln , subject to 1)( ˆ )( ˆ1 2 1 2 N n N nnxnx . To solve this problem let us consider the following Lagrangian 22 22 11 ˆˆˆˆ ln()()ln( )( ) NN nm Lxnxn xmx m 2 2 12 11 ˆˆ () 1() 1 NN nn xn xn . In order to simplify the computation, we set n pnx 2 )( ˆ and n pnx 2 )( ˆ. Hence we have 01ln1 n n p p L, 01ln 2 n n p p L, 1 1 N nn p , 1 1 N nn p . Solving the above equations, we finally obtain N nx 1 )( ˆ, N nx1 )( ˆ. From the definition of Shannon entropy, we know that if NxH ln ˆ and NxH ln ˆ , the equality in (3) holds. In addition, we also have )sin( N NN . From the proof we know that N nx 1 )( ˆ and N nx1 )( ˆ means that can be complex, and only if their amplitudes are constants, the equality will hold. Now we can obtain the following corollary out of above analysis. )( ˆ )( ˆnxandmx Corollary 1: For any given discrete time series )(,),3(),2(),1(,,,, 321 NxxxxxxxxX N CN with cardinality N and 1 2, ( ) is the DFRFT se- ries associated with the transform parameter X x ˆx ˆ ( , re- spectively), () counts the non-zero elements N N of (, respectively), if x ˆ x ˆ N nx 1 )( ˆ and N nx1 )( ˆ, then we have )sin( NNN and )sin(ln ˆˆ NxHxH . 3.2. Rényi Entropic Principle Theorem 2: For any given discrete time series )(,),3(),2(),1(,,,,321NxxxxxxxxX N CN with cardinality N and 1 2, ( ) is the DFRFT se- ries associated with the transform parameter X x ˆx ˆ ( , re- spectively), ( N) counts the non-zero elements of ( , respectively), then we can obtain the generalized discrete Renyi entropic uncertainty relation N x ˆx ˆ )sin(ln ˆˆ NxHxH with 1 2 1 and 2 11 . (4) where Copyright © 2013 SciRes. JSIP
Discrete Entropic Uncertainty Relations Associated with FRFT 123 N mmxxH 1 2 )( ˆ ln 1 1 ˆ , N nnxxH 1 2 )( ˆ ln 1 1 ˆ , which are Rényi entropies. Proof: In lemma 1, set 2q and 2p, we have 1 2 1 and 2 11 . Then from lemma 1, we obtain 2 1 1 2 2 1 2 1 1 2)( ˆ )sin()( ˆ N n N mnxNmx . Take the square of the above inequality, we have 1 1 2 1 1 1 2)( ˆ )sin()( ˆ N n N mnxNmx . Take the power 1 of both sides in above ine- quality, we obtain 1 1 1 2 1 1 1 1 2)( ˆ )sin()( ˆN n N mnxNmx , i.e., 1 )( ˆ )( ˆ )sin( 1 1 1 2 1 1 1 2 1 N m N n mx nxN . (5) Take natural logarithm in both sides of (5), we can obtain 22 11 11 ˆˆ ln( )ln() 11 NN nm xn xm ln sin()N . Clearly, as 1 and 1 , the Renyi entropy reduces to Shannon entropy, thus the Renyi entropic un- certainty relation in (4) reduces to the Shannon entropic uncertainty relation (3). Hence the proof of equality in theorem 2 is trivial according to the proof of theorem 1. Note that although Shannon entropic uncertainty rela- tion can be obtained by Rényi entropic uncertainty rela- tion, we still discuss them separately in the sake of inte- grality. 4. Resolution Analysis in Time-Frequency Domain In many cases [13-15], we often discuss the data concen- tration in both time domain and frequency domain, therefore, we will adopt a new measure on entropy: xHpxpHHpˆ )1( ˆ , where ( ) is the DFRFT series associated with the transform parameter x ˆx ˆ ( , respectively) for time series )(,),3(),2(, ),1(,,, 3 x 21xNxxxxxXNx, xH ˆ and x ˆ x ˆ H are the Shannon entropies for the DFRFT series (). Theorem 3: For any given discrete time series x ˆ )(,),3(),2(),1(,,,, ~ 321NxxxxxxxxX N with cardinality N and 1 2, ( ) is the DFRFT se- ries associated with the transform parameter X x ˆx ˆ ( , re- spectively), ( N) counts the non-zero elements of ( , respectively), then we can obtain the generalized discrete entropic uncertainty relation N x ˆx ˆ ))sin(ln( 2 1 NH , (6) Proof: For variable , we have for x ˆ p0 p N n p pnxx 1 1)( ˆˆ . Set )( ˆ max ˆnxx Nn . Therefore, we have q q q N nq q q qx nx x nx x dq d 1 1 1 11 1 1 1ˆ )( ˆ log ˆ )( ˆ ˆ log , (7) 2 1 2ˆ log q d dq 2 111 1111 11 1111 ˆˆˆˆ ()()()() 1log log ˆˆˆˆ qqq NN qqq nn qqq xnxnxn xn txxxx 1 q q q . Taking into account of the non-negative second de- rivative, therefore we know that q x1 ˆ log is convex. Therefore, we obtain 11 1 2 ˆˆ ˆ loglim log q qq q dd Hx xx dq dq log sin()N . (8) Since 1 ˆˆ 2 2 2 Xxx and ˆ 1 1ˆ sin( ) N and the Riesz-Thorin theorem [4], one has q q qN x x 2 1 1 1 1 )sin( ˆ ˆ 0 ˆ ,1 2 1 xq By applying negative logarith m to both sides of above inequality, we obtain the following relation 1 11 ˆˆ loglog q q xx 1log sin() 2 qN , 1 2 1q. Taking into account of 1 ˆˆ 2 2 2 Xxx , then both sides of above inequality are equal to zero for 2 1 q. Hence, (7) and (8) imply )sin(ln 2 1 ˆ 2 1 ˆ 2 1 NxHxH. Thus the proof is completed. Note that this proof can Copyright © 2013 SciRes. JSIP
Discrete Entropic Uncertainty Relations Associated with FRFT Copyright © 2013 SciRes. JSIP 124 be obtained via the similar manner with Section 3 A, however, we still give a new proof so that we can under- stand this point out of different aspect. On the other hand, different proofs yield the same result, as validates the conclusion derived here. 5. Conclusions In this article, we extended the entropic uncertainty rela- tions in DFRFT domains. We first introduced the gener- alized discrete Hausdorff-Young inequality. Based on this inequality, we derived the discrete Shannon entropic uncertainty relation and discrete Rényi entropic uncer- tainty relation. Interestingly, when the variable’s ampli- tude is equal to the con stant, i.e. the inverse of the squ are root of number of non-zero elements, the equality holds in the uncertainty relation. In addition, the product of the two numbers of non-zero elements is equal to )sin( N, i.e., )sin( NNN . REFERENCES [1] R. Ishii and K. Furukawa, “The Uncertainty Principle in Discrete Signals,” IEEE Trans Circuits and Systems, Vol. 33, No. 10, 1986, pp. 1032-1034. [2] L. C. Calvez and P. Vilbe, “On the Uncertainty Principle in Discrete Signals,” IEEE Trans Circuits and Systems-II: Analog and Digital Signal Processing, Vol. 39, No. 6, 1992, pp. 394-395. doi:10.1109/82.145299 [3] S. Shinde and M. G. Vikram, “An Uncertainty Principle for Real Signals in the Fractional Fourier Transform Do- main,” IEEE Transactions Signal Processing, Vol. 49, No. 11, 2001, pp. 2545-2548. doi:10.1109/78.960402 [4] G. L. Xu, X. T. Wang and X. G. Xu, “Generalized En- tropic Uncertainty Principle on Fractional Fourier trans- form,” Signal Processing, Vol. 89, No. 12, 2009, pp. 2692-2697. doi:10.1016/j.sigpro.2009.05.014 [5] R. Tao, B. Deng and Y. Wang, “Theory and Application of the Fractional Fourier Transform,” Tsinghua Univer- sity Press, Beijing, 2009 [6] S.C. Pei and J. J. Ding, “Eigenfunctions of Fourier and Fractiona l Fourier Transforms Wit h Complex Offsets and Parameters,” IEEE Trans Circuits and Systems-I: Regular Papers, 2007, Vol. 54, No. 7, pp. 1599-1611. [7] T. M. Cover and J. A. Thomas, “Elements of Information Theory, Second Edition,” John Wiley &Sons, Inc.,2006. [8] H. Maassen, “A Discrete Entropic Uncertainty Relation,” Quantum Probability and Applications, V, Springer-Verlag, New York, 1988, pp. 263-266. [9] C. E. Shannon. “A Mathematical Theory of Communica- tion,” The Bell System Technical Journal, Vol. 27, 1948, pp. 379-656 [10] A. Rényi, On Measures of Information and Entropy, In: Proceedings of the Fourth Berkeley Symposium on Mathematics, Statistics and Probability, 1960, p. 547. [11] G. Hardy, J. E. Littlewood and G. Pólya Inequalities. 2nd edition, Press of University of Cambridge, 1951. [12] D. Amir, T. M. Cover and J. A. Thomas, “Information Theoretic Inequalities,” IEEE Trans Information Theory, Vol. 37, No. 6, 2001, pp. 1501-1508. [13] D. Victor, O. Murad and P. Tomasz, “Resolution in Time–Frequency,” IEEE Transactions on Signal Proc- essing, Vol. 47, No. 3, 1999, pp. 783-788. doi:10.1109/78.747783 [14] T. Przebinda, V. DeBrunner and M. Özaydın, “The Op- timal Transform for the Discrete Hirschman Uncertainty Principle,” IEEE Transactions on information theory, Vol. 47, No. 5, 2001, pp. 2086-2090. doi:10.1109/18.930948 [15] X. D. Zhang, “Modern Signal Processing, “second edition) Tsinghua university press, Beingjing, 2002. [16] G. L. Xu, X. T. Wang and X. G. Xu, “The Logarithmic, Heisenberg’s and Short-time Uncertainty Principles As- sociated with Fractional Fourier Transform,” Signal Proc- essing, Vol. 89, No. 3, 2009, pp. 339-343. doi:10.1016/j.sigpro.2008.09.002
|