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

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

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

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

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

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

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

DFRFT by S method of a rectangular window function. The complexity of . MAY 1999 Fig. 6. k = 032 1 1 1 32. 5. 1. This concept is also applied in the time domain . 7. and x (n ) = 0 otherwise. the continuous FRFT in Fig. x(n) = 1 when 06 n 6. 4.1342 IEEE TRANSACTIONS ON SIGNAL PROCESSING. DFRFT by GSA method of a rectangular window function. When = 3=7. closer output to the continuous FRFT than Fig. an impulse-like output is obtained. ( and the rotation angle are When the number of points determined. DFRFT by OPA method of a rectangular window function. x (n ) = 1 when 06 n 6. and x (n ) = 0 otherwise. and x(n) = 0 otherwise. and it is the same as in the DFT case. NO. The output is close to Fig. If the rotational angles are adjusting. This ﬁgure has a closer Fig. the computation of the DFRFT can be implemented only by a transform kernel matrix multiplication. DFRFT by GSA of a chirp signal. the following method for implementing the DFRFT can be applied: (38) (39) frequency. 5. x (n ) = 1 when 06 n 6. 4. 47. 4. and the eigenvalues of DFRFT are also determined. VOL. output to the continuous FRFT than Fig. Then. computing the DFRFT is . the DFT Hermite eigenvectors can be computed a priori. identity transform) and any angle transform domains. This ﬁgure has a Fig. ej 20:1141k .

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

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

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

47. and the post-phase compensa- is the transform kernel of elliptical rotation. (52) (53) is the transform kernel of circular rotation. 13. MAY 1999 Fig. Three different cases of rotation concept for DFRFT in continuous time-frequency plane. • Elliptical Rotation: where (54) (55) (56) spread variance changes in these two schemes. is the elliptical rotation . 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. the circular rotation is just the FRFT of the signal. (59) . (57) where tion factor is equal to . VOL. the elliptical rotation will of signal become the circular rotation. If the Hermite eigenvectors in the circular rotation are from order 0 to inﬁnity. 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. The transform kernels of the circular and elliptical rotations deﬁned above are inﬁnite numbers of sums for Hermite eigenvectors.1346 IEEE TRANSACTIONS ON SIGNAL PROCESSING. is the circular rotation of signal . (a) (b) (c) Fig. NO. 5. 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. and where is deﬁned as in (43). While .

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

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

- 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

Sign up to vote on this title

UsefulNot useful- DSP , DAIICT by PRof Aditya Tatu,
- Fractional-Fourier-Transform.pdf
- Lie Gonzalez Ch5
- Image Transform 2
- DFT-Discrete Fourier Transform
- DFT
- Exam II Inel5309 f06
- FFT reference Sheets
- ch3.pdf
- Reviewof the Fourier Transform
- Lab Manual of DSIP
- fourier
- Accerelate Framwork
- ece6540_hw2
- dsp
- Lec-13
- Quiz 2013
- Project 2013
- Dtmf Decoding
- VL1
- signal and system
- Lab 8 - DFT Leakage and Windowing
- Dsp Endsem
- slika
- Green Sheet EE 210
- dsp qb_opt
- Brief Notes on Signals and Systems 7.2
- i i 3214921497
- 001
- eee
- 00757221 Discrete Fractional Fourier Transform Based on Orthogonal