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

BooksAudiobooksComicsSheet Music### Categories

### Categories

### Categories

### Publishers

Editors' Picks Books

Hand-picked favorites from

our editors

our editors

Editors' Picks Audiobooks

Hand-picked favorites from

our editors

our editors

Editors' Picks Comics

Hand-picked favorites from

our editors

our editors

Editors' Picks Sheet Music

Hand-picked favorites from

our editors

our editors

Top Books

What's trending, bestsellers,

award-winners & more

award-winners & more

Top Audiobooks

What's trending, bestsellers,

award-winners & more

award-winners & more

Top Comics

What's trending, bestsellers,

award-winners & more

award-winners & more

Top Sheet Music

What's trending, bestsellers,

award-winners & more

award-winners & more

Welcome to Scribd! Start your free trial and access books, documents and more.Find out more

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

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

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

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

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

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

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

4. The output is close to Fig. an impulse-like output is obtained. and x (n ) = 0 otherwise. This concept is also applied in the time domain . and x(n) = 0 otherwise. 1. If the rotational angles are adjusting. closer output to the continuous FRFT than Fig. 5. The complexity of . the computation of the DFRFT can be implemented only by a transform kernel matrix multiplication. k = 032 1 1 1 32. the following method for implementing the DFRFT can be applied: (38) (39) frequency. identity transform) and any angle transform domains. x(n) = 1 when 06 n 6. ej 20:1141k . 4. x (n ) = 1 when 06 n 6. and the eigenvalues of DFRFT are also determined. This ﬁgure has a Fig. MAY 1999 Fig. 4. 47. Then. the DFT Hermite eigenvectors can be computed a priori. and x (n ) = 0 otherwise. 6.1342 IEEE TRANSACTIONS ON SIGNAL PROCESSING. 7. When = 3=7. NO. This ﬁgure has a closer Fig. DFRFT by OPA method of a rectangular window function. the continuous FRFT in Fig. DFRFT by GSA method of a rectangular window function. and it is the same as in the DFT case. 5. computing the DFRFT is . output to the continuous FRFT than Fig. VOL. DFRFT by GSA of a chirp signal. x (n ) = 1 when 06 n 6. ( and the rotation angle are When the number of points determined. DFRFT by S method of a rectangular window function.

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

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

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

(a) (b) (c) Fig. Three different cases of rotation concept for DFRFT in continuous time-frequency plane. 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. 5. is the circular rotation of signal . • Elliptical Rotation: where (54) (55) (56) spread variance changes in these two schemes. 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 the post-phase compensa- is the transform kernel of elliptical rotation. NO. 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 elliptical rotation will of signal become the circular rotation. The transform kernels of the circular and elliptical rotations deﬁned above are inﬁnite numbers of sums for Hermite eigenvectors. is the elliptical rotation . 47. (57) where tion factor is equal to . While .1346 IEEE TRANSACTIONS ON SIGNAL PROCESSING. MAY 1999 Fig. and where is deﬁned as in (43). 12. (52) (53) is the transform kernel of circular rotation. (59) . VOL. the circular rotation is just the FRFT of the signal. 13. If the Hermite eigenvectors in the circular rotation are from order 0 to inﬁnity.

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

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

Download

Are you sure?

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

CANCEL

OK

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

Get the full title to continue

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

Restart preview

scribd