This action might not be possible to undo. Are you sure you want to continue?

5, MAY 1999

1335

**Discrete Fractional Fourier Transform Based on Orthogonal Projections
**

Soo-Chang Pei, Senior Member, IEEE, Min-Hung Yeh, and Chien-Cheng Tseng, Member, IEEE

Abstract—The continuous fractional Fourier transform (FRFT) performs a spectrum rotation of signal in the time–frequency plane, and it becomes an important tool for time-varying signal analysis. A discrete fractional Fourier transform has been recently developed by Santhanam and McClellan, but its results do not match those of the corresponding continuous fractional Fourier transforms. In this paper, we propose a new discrete fractional Fourier transform (DFRFT). The new DFRFT has DFT Hermite eigenvectors and retains the eigenvalue-eigenfunction relation as a continous FRFT. To obtain DFT Hermite eigenvectors, two orthogonal projection methods are introduced. Thus, the new DFRFT will provide similar transform and rotational properties as those of continuous fractional Fourier transforms. Moreover, the relationship between FRFT and the proposed DFRFT has been established in the same way as the conventional DFT-tocontinuous-Fourier transform. Index Terms— Discrete Fourier transform, discrete fractional Fourier transform, Fourier transform, fractional Fourier transform.

I. INTRODUCTION HE FOURIER transform (FT) is one of the most frequently used tools in signal analysis [1]. A generalization of the Fourier transform—the fractional Fourier transform (FRFT)—has been proposed in [2] and [3] and has become a powerful tool for time-varying signal analysis. In time-varying signal analysis, it is customary to use the time-frequency plane, with two orthogonal time and frequency axes [4]. Because the successive two forward Fourier transform operations will result in the reﬂected version of the original signal, the FT can in the be interpreted as a rotation of signal by the angle time–frequency plane and represented as an orthogonal signal representation for sinusoidal signal. The FRFT performs a rotation of signal in the continuous time–frequency plane to any angle and serves as an orthonormal signal representation for the chirp signal. The fractional Fourier transform is also called rotational Fourier transform or angular Fourier transform in some documents. Besides being a generalization of the FT, the

Manuscript received December 2, 1996; revised February 19, 1998. This work was supported by the National Science Council, R.O.C., under Contract NSC 85-2213-E002-025. The associate editor coordinating the review of this paper and approving it for publication was Dr. Akram Aldroubi. S.-C. Pei is with the Department of Electrical Engineering, National Taiwan University, Taipei, Taiwan, R.O.C. (e-mail: pei@cc.ee.ntu.edu.tw). M.-H. Yeh is with the Department of Computer Information Science, Tamsui Oxford University College, Tamsui, Taipei, Taiwan, R.O.C. C.-C. Tseng was with the Department of Electronics Engineering, Hwa Hsia College of Technology and Commerce, Taipei, Taiwan, R.O.C. He is now with the Department of Computer and Communication Engineering, National Kaohsiung First University of Science and Technology, Taipei, Taiwan, R.O.C. Publisher Item Identiﬁer S 1053-587X(99)03244-4.

T

FRFT has been proved to relate to other time-varying signal analysis tools, such as Wigner distribution [4], short-time Fourier transform [4], wavelet transform, and so on. The applications of the FRFT include solving differential equations [2], quantum mechanics [3], optical signal processing [5], timevariant ﬁltering and multiplexing [5]–[8], swept-frequency ﬁlters [9], pattern recognition [10], and time–frequency signal analysis [11]–[13]. Several properties of the FRFT in signal analysis have been summarized in [9]. Many methods for realizing the FRFT have been developed, but most of them are to utilize the optical implementation [14], [15] or numerical integration. Because the FRFT is a potentially useful tool for signal processing, the direct computation of the fractional Fourier transform in digital computers has become an important issue. The ideal discrete fractional Fourier transform (DFRFT) will be a generalization of the discrete Fourier transform (DFT) that obeys the rotation rules as the continuous FRFT and provides similar results as the FRFT. In [16], a method for a numerical integration FRFT has been proposed, but the method does not obey the rotation rules, and the signal cannot be recovered from its inverse transform. In [17], Santhanam and McClellan have developed a discrete FRFT, but their method does not provide the same transforms to match those of the continuous case. In this paper, we present a new discrete fractional Fourier transform (DFRFT). This DFRFT is a generalization of the DFT and will provide similar transforms as those of the continuous case. The relationship between the DFRFT and the FRFT can also be established and discussed in detail. Moreover, the proposed DFRFT has important unitary and rotation properties. This paper is organized as follows. In Section II, the previous development of continuous and discrete fractional Fourier transforms are reviewed. The concept for developing the DFRFT to have similar results as the continuous corresponding case are described in Section III. Two acceptable solutions for the DFRFT are considered and proposed in Section III. Then, the relationships between the FRFT and the DFRFT can be established in Section IV. Finally, conclusions and discussions are made in Section V. II. PRELIMINARY A. Continuous FRFT The Fourier transform of a signal can be interpreted as a angle rotation of the signal in the time–frequency plane. The FRFT is then developed and treated as a rotation of signal

1053–587X/99$10.00 © 1999 IEEE

However. The real parts of the FRFT or DFRFT in this paper are plotted by solid lines. 1. 1. 1 shows the FRFT of the rectangular window function [ for . Furthermore. and x(t) = 0 otherwise. a circular reﬂected version of the signal ( ). Imaginary part: dashed line. . and the English superscripts are used to denote the fractional power values of the DFT kernel matrices in this paper. the FRFT of the signal (4) (5) Equation (5) indicates that the FRFT can be interpreted as a weighting summation of Hermite functions. The conventional methods to compute the DFRFT in [17] and [20] are mainly on numerical computing the fractional power of the DFT kernel matrix. The weighting coefﬁcients are obtained from multiplying the phase term and the inner product of the input signal and the corresponding Hermite function. Fig. the (8) is the conventional DFT of signal . B. it will have a different phase term in the weighting coefﬁcients.1336 IEEE TRANSACTIONS ON SIGNAL PROCESSING. 2 shows the results of this DFRFT produced by a discrete rectangular window. . VOL. the middle is not the intermediate state of transform result for and . 5. Equation where is the linear (8) indicates that the DFRFT of the signal combination of the four major angular parts: the original signal . 47. The equivalence of (1) and (2) has been proved in [3]. The th-order normalized Hermite function with variance is deﬁned as (3) is the th-order Hermite polynomial [18]. to any angles in the time–frequency plane [9]. and a circular reﬂected version of its DFT . In order to avoid ambiguity. (2) provides an eigendecomposition representation of the FRFT kernel. MAY 1999 Fig. The discrete results shown in Fig. and ) are used to denote the rotation angles in the time–frequency plane. is the eigenfunction of the FRFT. 2 are quite different from the results of Fig. NO. the Hermite functions in [19] with a different scaling to those of the FRFT. otherwise here is deﬁned as [ ]. Real part: solid line. The rectangular window used . The fractional power of the DFT kernel matrix [20] is calculated by (6) (6) where (7) Applying the above deﬁned kernel to the signal is computed as DFRFT of the signal . and the rotation output of the signal is also the weighting summation of Hermite functions. Using the kernel of the by angle is computed as FRFT. Because where the normalized Hermite function with unitary variance. Fig. The Old DFRFT The kernel matrix of the DFRFT is obtained by computing the fractional power for the DFT kernel matrix. Fractional Fourier transform of a rectangular window. x(t) = 1 when 02 t 2. and the imaginary parts of the FRFT or DFRFT are indicated by dashed lines. The fractional power of the DFT kernel matrix and the rotation angles in the DFRFT essentially mean the same thing. [6]. The transform kernel of continuous FRFT is deﬁned as [2]. [9] csc if if if is not a multiple of (1) is a multiple of is a multiple of (2) where indicates the rotation angle of transformed signal for is the th-order normalized Hermite function the FRFT. the Greek subscripts (for example. In [19]. Thus. with unit variance. its DFT . a rotation operation using Hermite functions has also been proposed. elsewhere] for various angles. the two results . [3]. and .

is deﬁned. a method for computing the DFT eigenvectors has been introduced. Proposition 3: DFT Hermite eigenvectors should have the . and x (n ) = 0 otherwise. B. 1. but they correspond to different eigenvalues. Proposition 2: A matrix can be used to compute the real eigenvectors of the DFT matrix . and —and each is major eigensubspaces— . Matrix commutes with the DFT kernel . where is the spread variance of the Proof: It is assumed that DFT eigenvectors. and then. and matrix is deﬁned as OF DFRFT The continuous FRFT has a Hermite function with unitary variance as its eigenfunction. 2. . Old DFRFT of a rectangular window function. Eigendecomposition of DFRFT The development of our DFRFT is based on the eigendecomposition of the DFT kernel. it satisﬁes the commutative property (10) will also be the eigenvectors The eigenvectors of matrix of the DFT kernel matrix . . The eigenvectors of the DFT matrix constitute four . Proof: See [20]. Proof: See [21]. The continuous approximate form can be written as (13) Sampling by . x(n) = 1 when 06 n 6.PEI et al. The parameter is the index for the DFT eigenvalue . we get (12) is the th-order Hermite polynomial. . . The eigenvalue multiplicities of DFT matrix indicate the ranks of the DFT eigensubspaces [21]. The corresponding eigenfunction property for the DFT would be like (11) is the eigenvector of DFT corresponding to the where th-order discrete Hermite function. DEVELOPMENT A. . In [22]. (13) will become (14) .. . we have known that the eigendecomposition of the DFT kernel matrix is not unique. we only summarize some results for our further development of DFRFT. . Can the DFT have the eigenvectors with the similar shapes as the Hermite functions? These DFT eigenvectors are known as DFT Hermite eigenvectors in this paper. [22]. Proposition 1: The eigenvalues of and its multiplicities are shown in Table I. In [20]. . a multiplicity function function is used to denote the DFT eigenvalue multiplicity for . . . DFT Hermite Eigenvectors III. In the discussion in the previous subsection.: DISCRETE FRACTIONAL FOURIER TRANSFORM BASED ON ORTHOGONAL PROJECTIONS 1337 Fig. and many properties of the DFT matrix eigenvalues and eigenvectors have been discussed in [20] and [21]. If the Hermite function are sampled in this way. and corresponding to one of the four eigenvalues—1. a novel matrix is introduced to compute the real and complete set of DFT eigenvectors very elegantly. This In Table I. . (9) The Fourier transform of (13) can be computed as (15) . 1. but it cannot obtain the real-valued DFT eigenvectors. . where associated continuous spread variance is the sampling intervals of signal. —respectively. Here. . . from the continuous FRFT in Fig. are . . The output is quite different TABLE I MULTIPLICITIES OF THE EIGENVALUES OF DFT KERNEL MATRIX where matrix .

frequency spectrum that DFT can cover is the continuous spectrum spacing (or resolution) of the DFT is . Thus. the is. and our goal is to adjust the variance have the same eigenvector shape as itself. because larger is . we have the approximation expression (19) is wide This approximation is valid because is for large .1338 IEEE TRANSACTIONS ON SIGNAL PROCESSING. the sampling interval we will prove that even such samplings of Hermite functions can only have approximate Hermite eigenvectors of the DFT. Thus. . When the value of approaches inﬁnity. Next. the decay rate the degree of Hermite polynomial is proportional to for of the Hermite function sufﬁciently large . the larger order is. when order becomes large. Thus. then it can be shown that interval If is even (17) If is odd (18) for sufﬁciently large .e. For By truncating the integral interval of Fourier transform from to . However. the approximation in (17) becomes worse. approximated by can be From Proposition 3. Combining (19) and (20) very small when . is even. we only prove the case that odd. for (23) for then it can be shown that the DFT of the . it is clear that there are two approximation errors in (17). NO. VOL. and the decay rate of Gauss function very fast. The small aliasing will occur while is closer to .. both errors approach zero. taking at (21) Equation (14) becomes The proof is completed. we obtain and letting This expression is valid for any both sides. MAY 1999 The sampling theorem [23] tells us the continuous scope of . i. and the other is the numerical error in (20). and the samplings in Proposition 3 cannot guarantee that the numbers of the sign changes in this sampled th-order Hermite function are also . is obtained by samProposition 4: If the sequence pling the unit variance Hermite function with the sampling . we can ﬁnd that the DFT Hermite eigenvectors should have associate continuous spread variance . deﬁned in the range Proposition 5: If the sequence is obtained by shifting Hermite Gauss samples in the following way: If is even for (22) for and if is od. (16) adjusts the scaling of Hermite funcThe variance value to make the DFT tion. the better approximation (17) is. we have (20) (24) for sufﬁciently large . In Propositions 4 and 5. 5. Thus. The sequence obtained from (12) is also equal to the sampling of the Hermite function with the unit variance by . The th-order continuous Hermite function should have zeros [18]. by replacing the continuous integral with the numerical integral. The variable in (15) can be replaced by equal to to have its discrete form. 47. This implies that the truncation error in (19) is larger for high-order . but this will not inﬂuence the development of the DFRFT. One is the truncation error in (19). Proof: Here. the proof can be easily derived in the same way. The condition of for the two shapes of (14) and (16) being the same becomes This approximation is also valid because is is very large. these functions are not bandlimited. From the above proof. The sampled Hermite functions still can be used to construct the DFRFT kernel because they have the similar shapes and good approximations to the continuous Hermite functions. Next. . Therefore. we have . the slower decay rate the Hermite function will have.

it has been proved that the samplings of Hermite functions can have approximate DFT eigenvectors. the unit variance Hermite functions are in the following sampled with a period of discussions. we obtain DFT for for (30) The proof is completed. we also prove the case that is even. the choices for the DFT eigenvectors to construct the DFRFT kernel are multiple and not unique. The idea for developing our DFRFT is to ﬁnd the discrete form for (2). the terms of the Hermite functions are summed up from order zero to inﬁnity. we obtain C. This assignment rule matches the multiplicities of the eigenvalues of the DFT kernel matrix in Table I. The selections of the DFT Hermite eigenvectors are from low to high orders. Newly Developed DFRFT DFT (27) (28) is limited in the range where equality rewritten as DFT (29) is limited in the range where (28) and (29). The normalized vectors for the samplings of Hermite functions are deﬁned as (31) . Because of the above ambiguity. The While is given by DFT of the sequence DFT TABLE II EIGENVALUES ASSIGNMENT RULE OF DFRFT KERNEL MATRIX (25) Using the equality second term on the right side of (25) becomes . the DFT is computed from the eigenvectors of Hermite eigenvector matrix in the same DFT eigensubspace. In the case of continuous FRFT. we know that there are several DFRFT kernel matrices that can obey the rotation properties. In addition. Using the . only DFT Hermite eigenvectors can be added. • Ambiguity in Deciding the Fractional Powers of Eigenvalues: We know that the square root of unity are 1 and 1 from elementary mathematics. (26) Substituting (26) into (25) and using Proposition 4. In Propositions 4 and 5. we can compute in its DFT eigensubspace to obtain a the projections of Hermite-like DFT eigenvector (32) mod . Table II shows the eigenvalues assignment rules for the DFRFT. In order to retain the eigenfunction property in (11). for the discrete case. In addition. there exist two types of ambiguity in deciding the fractional power of the DFT kernel matrix.: DISCRETE FRACTIONAL FOURIER TRANSFORM BASED ON ORTHOGONAL PROJECTIONS 1339 Proof: Here. (27) can be The fractional power of matrix can be calculated from its eigendecomposition and the powers of eigenvalues. therefore. It is because the approximation error of the low DFT Hermite eigenvectors are small. will be a DFT Hermite eigenvector. Unfortunately. and is the eigenvector of matrix where . the proof can also be easily derived. . the Because matrix can have complete real orthogonal DFT eigenvectors. the eigenvectors can be used as bases for individual DFT eigensubspaces.PEI et al. In (32). • Ambiguity in Deciding the Eigenvectors of the DFT Kernel Matrix: The DFT eigenvectors constitute four major eigensubspaces. is odd. we should not expect that a ﬁnite vector can express the oscillation behavior of the very high-order Hermite function very well. Combining . eigenvectors for the However. This indicates that there exists root ambiguity in deciding the fractional power of eigenvalues.

(34a) . Unfortunately. the eigenvalue–eigenfunction relation is retained in deﬁning the DFRFT. This means that is replaced by in (34). . . The Gram–Schmidt Algorithm (GSA): The Gram–Schmidt method [29] is a well-known orthogonalization approach for vectors. . . whereas is odd. the eigenvectors obtained from matrix are directly considered to be the discrete Hermite functions. Therefore. and for even . shown at . In [25]. the convergence for the eigenvectors obtained from matrix are not so fast for the high-order Hermite functions by the LMS error criterion. MAY 1999 The transform kernel of the DFRFT can be deﬁned as (33) for (odd) (34) for (even) where . . is the normalized eigenvector corresponding to the th-order Hermite function. The DFT eigenvectors are orthogonalized for each eigensubspace. . This is because the eigenvectors located in different eigensubspaces will be always orthogonal. the eigenvectors obtained from the matrix are just discrete Mathieu functions [21]. So the DFT Hermite eigenvectors can be orthogonalized for every eigensubspace individually to obtain orthogonal eigenvectors in the whole eigenspace of DFT.. Calculate the continuous samples of Hermite functions: Compute the eigenvectors of matrix : Using equation (32) to compute Hermite eigenvectors by projections: for to for to end end . Example 1— : The rotation angle is equal to . Equation (32) provides a method for ﬁnding DFT Hermite eigenvectors. Although the Mathieu functions can converge to a Hermite function [28]. Two methods for vector orthogonalization are proposed in the following of this paper. the transform kernel for the rotation angle can be computed as in (34a). However. is the bottom of the page. is deﬁned as follows: where For odd . nonorthogonalized Hermite eigenvector. orthogonalized Hermite eigenvector. In addition. It is easy to verify that the angle rotation property of the DFRFT can be preserved only while the DFT eigenvectors are orthogonal.. The role of matrix in (32) is just as a tool to ﬁnd a complete set of real and orthogonal DFT eigenvectors. for and the eigenvalues of the DFRFT are .1340 IEEE TRANSACTIONS ON SIGNAL PROCESSING. vector orthogonalization is required for the DFT Hermite eigenvectors obtained from (32). 5. . where the eigenvector corresponding to the th continuous Hermite function. VOL. Therefore. . The symbol notation in developing the two algorithms are as follows: continuous Hermite samples vector. the DFT Hermite eigenvectors obtained from (32) cannot constitute an orthogonal basis for DFT eigenspace. Such a method is called the method in this paper. 47. . . An eigenbased method for computing the DFRFT to have similar continuous results has been proposed by us in [24]–[27]. NO. and for is even. It is easy to show that the eigenvectors located in different eigensubspaces will be orthogonal. The DFT Hermite eigenvectors in each DFT eigensubspace can be orthogonalized by the Gram–Schmidt method. It must be mentioned that the last eigenvalue and should be assigned to the is not assigned to to match the eigenvalue multiplicities value of the DFT matrix in Table I.

and will can be found by be our solution. it can be to found that the transform results change from a chirp signal ) to an impulse-like function ( ). 8–10. Because the . satisfy Algorithm Calculate the continuous samples of Hermite functions: Compute the eigenvectors of : Using equation (32) to compute Hermite eigenvectors by projections: for to Compute the SVD of . Orthogonal Procrustes Algorithm (OPA): A traditional mathematical problem known as the orthogonal Procrustes Algorithm can be used to ﬁnd the least Frobenius norm [30. is the subject to (36) the . where . otherwise. The sampled discrete data then becomes . and OPA. The error vectors between samples of Hermite functions and DFT Hermite eigenvectors are deﬁned as (37) Example 2: In this example. 7. GSA. [ which is the same as the signal used in Fig. The corresponding samples of the continuous FRFT for the impulse signal are plotted in Fig. the solution will also . 582] for the two given spaces. we use the DFRFT to deal with the rectangular window function shown in Fig. . . and the number of points sampling interval is equal to 73. ) are plotted Transform results of the impulse signal ( in Figs. 2. ( the DFRFT can be used for the chirp signal and chirp rate detection. 8–10 for the method. Frobenius norm of (35) samples of Hermite functions are plotted for . Figs. In Fig. respectively. 3. The continuous FRFT is an orthonormal signal decomposition for chirp signals [9]. The chirp signal used here is equal . . the norms of error vectors between the computed DFT Hermite eigenvectors and . In Fig. 3. and the OPA minimizes the total errors between the samples of Hermite functions and orthogonal DFT Hermite eigenvectors. Based on the unitary property in the DFRFT and the transform results shown in Figs. we can ﬁnd that the proposed DFRFT provides a similar orthonormal signal decomposition for discrete chirplike signals. Therefore. 1. 11. for minimize minimize where matrix. p. Norms of error vectors between DFT Hermite eigenvectors and samples of Hermite functions.: DISCRETE FRACTIONAL FOURIER TRANSFORM BASED ON ORTHOGONAL PROJECTIONS 1341 Fig. The is equal to 4/13.PEI et al. The minimizing calculating the singular value decomposition (SVD) of . D. E. Implementation of the New DFRFT As in the case of DFT frequency domain. Example 3: The DFRFT by the GSA for a chirp signal is computed in this example. Properties of DFRFT The properties of the DFRFT are shown in Table III. A more detailed theory and algorithm can be found in [7]. 4–6 show the DFRFT of the rectangular window function for various angles for method. the last half of the indices in the DFRFT must also be treated as the negative end The GSA minimizes the errors between the samples of Hermite functions and orthogonal DFT Hermite eigenvectors from low to high orders. We can formulate our problem as the OPA to ﬁnd the least Frobenius norm between the samples of Hermite functions and orthogonal DFT Hermite eigenvectors. GSA. 9 and 10. and OPA. respectively. The norm of the error vectors between the DFRFT and samples of the FRFT are also shown in the titles of Figs. ].

the DFT Hermite eigenvectors can be computed a priori. 4. 5. and x(n) = 0 otherwise. computing the DFRFT is . ( and the rotation angle are When the number of points determined. DFRFT by OPA method of a rectangular window function. k = 032 1 1 1 32. 7. the computation of the DFRFT can be implemented only by a transform kernel matrix multiplication. 1. x (n ) = 1 when 06 n 6. DFRFT by S method of a rectangular window function. the continuous FRFT in Fig. x(n) = 1 when 06 n 6. 5. ej 20:1141k . NO. The complexity of . output to the continuous FRFT than Fig. This ﬁgure has a Fig. and x (n ) = 0 otherwise. an impulse-like output is obtained. closer output to the continuous FRFT than Fig. x (n ) = 1 when 06 n 6. and x (n ) = 0 otherwise. MAY 1999 Fig. identity transform) and any angle transform domains.1342 IEEE TRANSACTIONS ON SIGNAL PROCESSING. VOL. the following method for implementing the DFRFT can be applied: (38) (39) frequency. When = 3=7. 4. This concept is also applied in the time domain . If the rotational angles are adjusting. The output is close to Fig. 4. 6. DFRFT by GSA of a chirp signal. and the eigenvalues of DFRFT are also determined. Then. 47. This ﬁgure has a closer Fig. DFRFT by GSA method of a rectangular window function. and it is the same as in the DFT case.

Here. This is the major reason for the mismatches in [17]. 6 (40) The deﬁnitions of matrix and matrix are the same as those . and x(n) = 0 otherwise. It can be seen that chirp-like outputs are obtained while the angle = =2. Fig. Proof: Let be any DFT eigenvector located in the . 6 Fig. we can obtain is also a DFRFT eigenvector. When = =2. x(0) = 1. When = =2. DFRFT by GSA of an impulse function. the inner products of signal and eigenvectors. and the output becomes a constant. Applying the transform eigensubspace kernel deﬁned in (6) to the eigenvector . A rigorous discussion for the mismatches of [17] has been presented in [16]. and x(n) = 0 otherwise. and the value is the eigenvalue for this eigenvector . 8. Since any eigenvector in Therefore. and the output becomes a constant. DFRFT by S method of an impulse function. and the output becomes a constant. . 10. the DFRFT reduces to DFT. x(0) = 1. and x(n) = 0 otherwise. only should be recomputed. 9. we stress the major reason for this mismatch. The coefﬁcients ’s are in (33). If the rotation angle is changed. It can be seen that chirp-like outputs are obtained while the angle = =2. the DFRFT reduces to DFT. x(0) = 1.PEI et al. Proposition 6: The method in (6) assigns all the eigenvectors of the DFRFT matrix to only four different eigenvalues. DFRFT by OPA of an impulse function. the diagonal matrix F. Discussion The method in [17] obeys the rotation properties. It can be seen that chirp-like outputs are obtained while the angle 6= =2. the DFRFT reduces to DFT. and they can be computed in advance. When = =2.: DISCRETE FRACTIONAL FOURIER TRANSFORM BASED ON ORTHOGONAL PROJECTIONS 1343 Fig. but it cannot have similar results as in the continuous case. .

In the conventional DFT analysis. X is even. Thus. b) = F (b. To begin with. It should be noted that the number of eigenfunctions for the continuous FRFT in (2) is inﬁnite. In the following discussion. x(0) = 1. RELATIONSHIP BETWEEN FRFT AND DFRFT In this section. VOL. and the frequency resolution is [23]. If x is odd. (41) is the eigenvector corresponding to the th-order where Hermite function. NO. 5. However. Transform Range and Resolution of DFRFT In this subsection.b)-element in DFRFT kernel matrix F [vn ] = e0jn vn where vn is Hermite-like function N F [(k)] 2 10j2cot ej cot If x is even. Then. the number of DFT Hermite eigenvectors is only ﬁnite. The the overall frequency range that the DFT can represent is equal to . 11. the method assigns the DFRFT eigenvalues to only four values. 47. the signal and transform results in [31] are discrete and periodic. the proposed DFRFT has assigned a different eigenvalue for each eigenvector. is the total recorded signal duration and is signal. Samples of continuous FRFT of an impulse function. MAY 1999 Fig. it is also is the number of points of the discrete the time resolution. It is based on retaining the property of the DFT that a sampled periodic function transforms into a periodic function. X is odd. In the DFRFT of [17]. it has been proved that the spread variance . and there are some approximation errors in the DFT Hermite approaches inﬁnity. we will discuss the transform range and resolution for the DFRFT (see Table IV). While errors of the DFT Hermite eigenvectors will be reduced. The DFRFT developed in this paper is based on the mimicking of the eigenvalue–eigenfunction relationship that the continuous counterpart of this transform has with the normalized unit variance Hermite functions. IV. In [31]. a) where F (a. Therefore. we will of the DFT Hermite eigenvector is compute the FRFT for a Hermite function with any variance . the periods in the transform results change for different rotation angles. there will be four different eigenvalues for all the eigenvectors. we will review and understand the transform range and resolution for the is the conventional DFT. The development of the new DFRFT is based on the same idea as (2). TABLE III PROPERTIES OF DFRFT Unitary Angle additivity Time Inversion Periodicity Symmetry Eigenfunction Impulse Transform Parity F )3 = (F )01 = F0 F F =F F x(0n) = X (0n) F =F F (a. . On the other hand.1344 IEEE TRANSACTIONS ON SIGNAL PROCESSING. sampling interval for the original continuous signal. the approximation eigenvectors. and more DFT Hermite eigenvectors are used to compute the DFRFT. A. Moreover. Here. our DFRFT can be used for any rotation angle and can provide very similar results as continuous cases. b) is the (a. and the rotation angle in [31] is valid for a certain discrete set of rotation angles. we will establish the relationship between the FRFT and the DFRFT. the DFRFT can be used to give the similar continuous transform results within the accuracy of the discrete ﬁnite vector approximation. and x(t) = 0 otherwise. the transform range and resolution of the DFT have been well discussed [23]. In Proposition 3. ( one of the four eigensubspace has the same eigenvalue. and transform range in the time domain as well. an alternative DFRFT has recently been proposed. which satisﬁes the eigenvalue and eigenfunction relationship as a continuous FRFT kernel.

whereas This means that the DFRFT is implemented in a circular rotation for the discrete case. B. The sampled spread variance of the DFRFT is vector in (12) can also be an approximate eigenvector in the fractional Fourier domain (angle ). 12. The symbol and are the two special cases. the DFRFT performs independent circular rotation in the discrete notation. the continuous time–frequency plane. From (45) and Fig. but the DFRFT performs elliptical rotation in the continuous time–frequency plane while . it has been shown that there are approximation errors in the DFT Hermite eigenvectors. the spread variance of the DFRFT for angle is equal to (47) The overall transform range the DFRFT can cover is equal to (50) In (34).PEI et al. 12. therefore. as drawn in Fig. In this subsection. For simpliﬁcation of analysis. Angle actual angle for the elliptical rotation. It is easy to check some special cases for (43): and . it can be found that the signal rotation of the DFRFT is an elliptical rotation in . the approximation errors of the DFT Hermite eigenvectors are ignored in the following discussion. we can substitute the variance in (43) by to get the spread variance of the DFRFT in the fractional Fourier domain (angle ) (48) Both (47) and (48) indicate the corresponding variance of DFT Hermite eigenvectors in the fractional Fourier domain. only the last term can affect the envelope . 12. frequency ( The rotation concept of the DFRFT in the continuous is the time–frequency plane is plotted in Fig. the resolution of the FRFT can be obtained as (49) (42) where puted as . Elliptical Rotation versus Circular Rotation The FRFT performs circular rotation in the continuous time–frequency plane. and is the where number of points for the discrete signal. The overall transform range of the DFRFT can cover is equal to (45) Proof: From Proposition 3. frequency ( • The length of the time range is shorter than that of the ). • The lengths of time and frequency ranges are equal ). Proposition 8: The resolution of the DFRFT with angular is equal to parameter (44) is the sampling interval of the signal. it has been known that the . will be equal to 1 for any value of . (46) Thus. • Circular Rotation: (51) . If . In (42). Three cases can be realized for the signal rotation of the DFRFT in the continuous time–frequency plane. is used to denote the resolution of the FRFT. but it actually performs elliptical rotation in the continuous time–frequency plane. we will deﬁne circular and elliptical rotations clearly and establish the relationship between circular and elliptical rotations. of the FRFT of From Propositions 4 and 5. The new variance is com- (43) Proof: This proposition can be easily proved by computing the FRFT of the normalized Hermite functions.: DISCRETE FRACTIONAL FOURIER TRANSFORM BASED ON ORTHOGONAL PROJECTIONS 1345 TABLE IV TRANSFORM RANGE AND RESOLUTION OF DFRFT Proposition 7: The continuous FRFT of a Hermite function with variance is equal to Moreover. ( • The length of the time range is longer than that of the ).

13. is the circular rotation of signal . 12. The critical point for which we care is the (58) can be written as the weighted Proof: The signal sum of the normalized Hermite functions. is the elliptical rotation . (52) (53) is the transform kernel of circular rotation. MAY 1999 Fig. If the Hermite eigenvectors in the circular rotation are from order 0 to inﬁnity. (57) where tion factor is equal to . (a) (b) (c) Fig. Results of Example 4 showing that the error norms between DFRFT and continuous FRFT samples are greatly reduced by angle modiﬁcation and post-phase compensation. Proposition 9: The circular rotation of signal can be implemented by an angle modiﬁcation of the elliptical rotation and post-multiplying a phase compensation factor. VOL. the circular rotation is just the FRFT of the signal. and the post-phase compensa- is the transform kernel of elliptical rotation. 5. While . and where is deﬁned as in (43). Three different cases of rotation concept for DFRFT in continuous time-frequency plane. 47. (59) . • Elliptical Rotation: where (54) (55) (56) spread variance changes in these two schemes. NO.1346 IEEE TRANSACTIONS ON SIGNAL PROCESSING. The transform kernels of the circular and elliptical rotations deﬁned above are inﬁnite numbers of sums for Hermite eigenvectors. the elliptical rotation will of signal become the circular rotation.

However. The discrete post-phase compensation resolution is still factor for the DFRFT is the length of time range ( ) will almost equal the length ). we only compute the results of by the GSA. the modiﬁcation of the angle and postphase compensation discussed above are very critical. and the number of points in the interval is still signal becomes 37. therefore. 13(c) shows the sample values of the continuous FRFT . whereas the OPA minimizes the total errors between the samples of the Hermite functions and the orthogonal DFT Hermite eigenvectors.: DISCRETE FRACTIONAL FOURIER TRANSFORM BASED ON ORTHOGONAL PROJECTIONS 1347 where . Compute the FRFT of (circular rotation) for the both sides of (59). . Fig. 13(a) shows the DFRFT with angular parameter for the original signal. Fig. Here. The correis equal to 1. Therefore. Step 3) Multiply the result obtained from Step 2 by the post-phase compensation factor shown in (62). The operation in the DFRFT performs elliptical rotation in the continuous time–frequency plane for the case . Step 2) Calculate the -point DFRFT of the signal with the angular parameter . Fig. Using Proposition 9. It must be noted that the postphase where compensation factors in (58) and (62) are different.PEI et al. and the transform . the elliptical rotation can be implemented by circular rotation. The length of the time range is almost equal to the length of the frequency range. Two vector orthogonalization processes for the DFT Hermite eigenvectors are accordingly proposed in this paper: One is GSA. If variance of the continuous-time counterpart of the eigenvector will approximate to unity. the sampling . we directly compute the DFRFT without angle modiﬁcation and postphase compensation. (62) . Example 4: In this example. 13(b) is the DFRFT with angular parameter modiﬁcation and postphase compensation. The new transform retains the eigenvalue–eigenfunction relationship using the sampled version of the normalized unit-variance Hermite functions that the continuous FRFT has with the unit variance Hermite functions. where . and the other is OPA. If the values of (57) are . we again deal with the rectangular window shown in Example 2. The rotation will almost be of the frequency range ( a circular rotation. With the help of the commutative matrix . such DFT Hermite eigenvectors cannot form an orthogonal basis for the DFT eigenspaces. The GSA minimizes the errors between the samples of the Hermite functions and the orthogonal DFT Hermite eigenvectors from low to high orders. The in (58) is replaced by in (62) for preserving variable . the DFRFT in Example 2 is almost a circular rotation. the the unitary property in the DFRFT. the following steps must be used to make the results of discrete cases match those of the continuous case. it will become evaluated at the points (60) (61) can be Equation (61) indicates that the FRFT with angle implemented by a DFRFT with angle . The DFT Hermite eigenvectors can be calculated by the projection of samples of the unit variance Hermite functions in the DFT eigensubspaces through the help of the eigenvectors of the matrix . and Fig. 13(a) is quite different from Fig. However. and the effects of angle modiﬁcation and postphase compensation will be very small. While the variance is far from unity. The rotation angle .0488 in Example sponding variance 2. In Example 2. V. for the indices . 13(c) very well. In this case. which are equal to continuous counterpart of the variance in this example is . 13(b) match the corresponding continuous FRFT cases shown in Fig. the complete real and orthonormal eigenvectors of the DFT kernel matrix can be computed. 13(c) due to the elliptical rotation and not in the circular rotation in the continuous time–frequency plane. The results shown in Fig. CONCLUSIONS The development of this DFRFT is based on the eigendecomposition of the DFT kernel matrix . Step 1) Compute the modiﬁcation angle .

Amer. Ozaktas. pp. Kutay. M. 7599–7602. A.D. optical information processing. C. vol. 1993. 1972. respectively. pp. 25. Shyu. The Fourier Transform and Its Applications. Applicat. 15–16. 1. Presently. and Ph. 6th IEEE DSP Workshop. Tatung Institute of Technology. “Fractional correlation. 42. June 1994. pp. degrees from the University of California. Yeh. “On Namias’ fractional Fourier transforms. If the DFRFT performs a circular rotation of the signal in the . vol. 3. He received the B. 119–124. degree in computer science and information engineering in 1992 and the Ph. 1959. vol. 474–478. in 1988. Kerr. 1986. Taipei. time–frequency analysis. pp. 665–675. Math. C.S. Mendlovic. 1993. 2141–2150. in 1949. Opt. W. Oppenheim. Taiwan. T. pp. Hsinchu. Lohmann. MD: Johns Hopkins Univ.. [25] 22. 1798–1801. Tamsui Oxford University College. Arikan. pp. and wavelets. “Eigenvalue and eigenvector decomposition of the discrete Fourier transform. M. 1191–1194. 1953. W. [24] S. 2nd ed. H. and D. New York: McGraw-Hill.. A. 10. [23] A. H. pp. His main research interests are in fractional Fourier transforms. [14] D.” IEEE Signal Processing Mag. O. Opt. [22] G. degree from the National Taiwan University (NTU).S. Barshan. Mar. June 1996. A. Taipei. [3] V. Taoyuan. Kutay.” J. 1994. Y. Ozaktas. F. in 1970 and the M. Bourdeaux-Bartels. 1996. Lohmann. pp. M.S. 5. vol. Mendlovic. B. vol. in 1990 and 1995. vol.E. H. pp. Soc. 1994. pp.. 47. Ozaktas.. NTU. W. [2] A. Signal Processing. Opt. [17] B. 10. “Eigenvectors and functions of the discrete Fourier transform. E. 1996. 39. R. 1982. Ozaktas. Mendlovic. [28] P. Namias. “Improved discrete fractional Fourier transform. “Separation of transient signals. Acoust. and electronic commerce. [7] R. “Convolution and ﬁltering in fractional Fourier domain.. Signal Processing. and M.-H. He received the B. W. [6] H. 1996. [4] F. Aademir. and J.C.” J. REFERENCES [1] R. [5] H.O.” in Proc. 1989. pp.. . Taiwan. A. Apr.” IEEE Trans. vol.” Appl. and G. W.” J. “The fractional order Fourier transform and its application to quantum mechanics. NO. 46. Dickinson and K. in 1972 and 1975. Cincotti. pp. Santarsiero. Santhanam and J. “Relationships between the Radon–Wigner and fractional Fourier transforms.. ASSP-30. Opt. VOL. from the Tatung Institute of Technology. [19] B. “Fractional Wigner distribution function. M. vol. “Convolution. 13. Symp.E.” Appl. F. he was a Research Assistant at UCSB. [15] . “Effect of fractional Fourier transformation on time-frequency distributions belonging to the Cohen class. Feschbach. National Kaohsiung First University of Science and Technology. Parks. Santa Barbara (UCSB).. “Discrete fractional Hartley and Fourier transforms. V. “Chirp ﬁltering in the fractional Fourier domain. Ozaktas. Almeida. Feb. 1995. Insel. 1996. pp. A. and R.” IEEE Trans. [8] A. Methods of Theoretical Physics. 994–998. The complexity for . N. and multiplexing in fractional Fourier domains and their relationship to chirp and wavelet transforms. “Fractional Fourier domains. pp. and O. Circuits Syst. Taiwan.C. vol. Apr. 1997. Taipei. pp.D. Taipei. Mar.. Yeh. From 1995 to 1997. New York. 44. Pei is a Member of Eta Kappa Nu and the Optical Society of America.. Dragonman. Amer.” IEEE Signal Processing Lett. [27] S. R.. 1416. pp.D. vol. Oct. This DFRFT provides a method for implementing the discrete FRFT. July 15. Orthogonal Functions. 1994. 1992. 205–207. degrees from the National Taiwan University. F. Taipei. 25. Taiwan. 33. degree in computer engineering from the National Chiao-Tung University. Englewood Cliffs. Amer. Hwa Hsia College of Technology and Commerce. Shenoy. Linear Algebra. he was an Associate Research Engineer at Telecommunication Laboratories. H. Chunghwa Telecom Company. K. 2522–2531. A. The DFRFT proposed in this paper not only supplies the similar transforms to match with those of the continuous case but also preserves the rotation properties.” J. Signal Processing. 1992. Mendlovic. [9] L. C. Taiwan. H. Matrix Computations. p. D.. Feb. Circuits Syst. and laser holography. [21] J. Bozdagi. Soc. ﬁltering.” Signal Process. Baltimore. He was a Professor and Chairman with the Electrical Engineering Department.. A. . Chien-Cheng Tseng (S’90–M’95) was born in Taipei. Gori. vol. “The fractional Fourier transform and time-frequency representation. Symp. Ozaktas. [12] D. II. “Generalized self-Fourier functions. A. He is currently an Assistant Professor with the Department of Computer and Communication Engineering. Morse and H. 536–539. Ozaktas. B. H. Signal Processing. 42. M. Speech. 1875–1881. pp. [31] O.” in Proc. M.” Opt. 66–74. all in electrical engineering. [13] H. A. 1987. pp. . Ozaktas. and D. Soc. degree in electrical engineering in 1997. His research interests include digital signal processing. He was an Engineering Ofﬁcer in the Chinese Navy Shipyard from 1970 to 1971. IEEE Int.. R. [11] H. May 1996. J. He is currently an Assistant Professor with the Department of Computer Information Science. From 1997 to 1998. Barshan. [30] G. Steiglitz. vol.E. “Discrete fractional Fourier transform. “Fractional Fourier transformations and their optical implementation—II. C. New York: McGraw-Hill. Onural. His research interests include digital signal processing. McBride and F. N. vol.” IMA J..” Opt. Pei. Pei and M. H. Sansone. “Two dimensional discrete fractional Fourier transform. pp. Time-Scale Anal. Parks. Dorsch. 67. vol. with honors.” IEEE Trans. Ltd. Soc. 9. 40–41. Taipei. 547–559. Press.S.” IEEE Trans.” J. vol. From 1971 to 1975. 303–309. Inst. 1989. degree. Interscience. [29] S. and L. and A. 25–31. Audio Electroacoust. and M. G. Feb.. Amer. AU-20. Soffer. 1047–1049. 21–67. “The discrete rotational Fourier transform. DFRFT can be established as follows. and it is an important tool for signal processing. Bitran. On the other hand. Kutay. 11. [20] B. Opt. M. Golub and C. he was an Assistant Professor of the Department of Electronics Engineering.E.S. pp. in 1965. “Digital computation of the fractional Fourier transform. Sept. Nov. Opt. IEEE Int. Discrete-Time Signal Processing. 1989. M. Tseng. Time–Freq.” J.” IEEE Trans. 199–203. Phys. Hlawatsch and G. which is the same as that implementing the DFRFT is of the DFT. Math.S. he is Professor and Chairman of the Electrical Engineering Department. Van Loan. and Ph.C. Soo-Chang Pei (SM’89) was born in Soo-Auo. pp. Opt. in 1964. if the DFRFT becomes an elliptical rotation in the continuous time–frequency plane. Soc. vol. Taipei. 34. McClellan and T. in 1987. 1980.” Signal [26] Process. H. pp. vol. McClellan. B.1348 IEEE TRANSACTIONS ON SIGNAL PROCESSING. Lett. 1998. Weisburn. Lohmann and B. Friedberg. 1995. Dr. 241–265. 99–108.. M. pp. vol. H. W. A. MAY 1999 Furthermore.” J.O. Amer. [16] H. Arikan. NJ: Prentice-Hall. both from the National Taiwan University. Tamsui. pp. [18] G.O. from 1981 to 1983. vol. An angle modiﬁcation and a postphase compensation in the DFRFT for elliptical rotation are required to match the results that are similar for the continuous FRFT. A. and L. vol. pattern recognition. He then received the M. Rev. Erkaya. and the M.” IEEE Trans. Mendlovic and H.” in Proc. NJ: Prentice-Hall.. M. Bracewell. Appl. Jan. pp. 1994.. Taipei. “Linear and quadratic timefrequency signal representations. respectively. the relationship between the FRFT and the . Min-Hung Yeh was born in Taipei. 1994. 3084–3091. vol. time–frequency plane.-C. G. image processing. 1998. He received the B. M.. 11. 45. Spence. 159–175. [10] D. J. “The discrete fractional Fourier transform. M. Englewood Cliffs. vol. “Fractional Fourier transformations and their optical implementation—I.

- 3 Equality
- Venus & Adonis - Shakespeare
- Troilus & Cressida - Shakespeare
- Tempest - Shakespeare
- Hamlet by William Shakespeare
- Phoenix & Turtle - Shakespeare
- Passionate Pilgrim - Shakespeare
- Othello - Shakespeare
- Merry Wives of Wndsor - Shakespeare
- Merchant of Venice - Shakespeare
- Lucrece - Shakespeare
- Love's Labours Lost - Shakespeare
- Julius Caeser - Shakespeare
- Comedy of Errors - Shakespeare
- As You Like It - Shakespeare
- Antony & Cleopatra - Shakespeare
- All is Well That Ends Well - Shakespeare
- UBUNTU8.04TUT
- PC Magazine - Linux Solutions Malestrom
- Linux Sys Calls
- Linux Drivers
- Linux Unix Concepts
- Intro Linux Device Drivers
- Embedded Linux Kernel and Drivers
- Arcom Embedded Linux Technical Manual

- Discrete Rotational Fourier Transform
- Fractional-Fourier-Transform.pdf
- Image Transform 2
- The DFrFT
- Helin-useful.pdf
- 07343744
- Discrete Fourier transform - Wikipedia, the free encyclopedia.pdf
- Discrete Fourier Transform
- UNIT 2 DIGITAL IMAGE PROCESSING
- Image Transforms
- ch3.pdf
- DSP_Chapter03.ppt
- Digital Signal processing Chapter 3
- Dsp technique
- Discrete sine transform
- Linear Systems
- Chapter 04
- Short Time Fourier Transform
- DSP & Digital Filters
- experimentno-131231022848-phpapp01
- Rr410201 Digital Signal Processing
- STFT
- Quiz 2013
- Matlab Dft
- Dsp for Matlab and Labview
- Performance Comparison of Image Retrieval Using Fractional Coefficients of Transformed Image Using DCT,WASH,HAAR and Kekre's Transform
- Discrete Fourier Transform
- DIGITAL IMAGE WATERMARKING USING DFT ALGORITHM
- Transform Based Watermarking
- Least Squares

Are you sure?

This action might not be possible to undo. Are you sure you want to continue?

We've moved you to where you read on your other device.

Get the full title to continue

Get the full title to continue reading from where you left off, or restart the preview.

scribd