7 L=3,Γ 1 =4,Γ 2 =5,Γ 3 =7, relative errors 3.5 x 10−3 maximal remainder error level τ 3 estimation error <strong>and</strong> its upper bound of n 2.5 2 1.5 1 0.5 estimation error of n divided by the mean of n error upper bound divided by the mean of n 0 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 Fig. 3. Relative estimate error of ˆn in (23) <strong>and</strong> its upper bound in (24) using proposed robust CRT. APPENDIX: PROOF OF THEOREM 1 From Condition (16) on x, it is not hard to see that the true solution n i in (8) falls in the range 0 ≤ n i < γ i for 1 ≤ i ≤ L. Thus, for 2 ≤ i ≤ L <strong>and</strong> any (¯n 1 , ¯n i ) ∈ S i , we have ∣ ∣¯n iΓ i + ˜k i Γ i M − ¯n 1Γ 1 − ˜k ∣ 1 Γ ∣∣∣∣ 1 M ∣ ≤ n i Γ i + ˜k i Γ i M − n 1Γ 1 − ˜k 1 Γ 1 M ∣ . (27) From (8), Γσx = n i Γ i + ˜k i M Γ i + ǫ i Γ i , 1 ≤ i ≤ L. (28) Let µ i ∆ = ¯ni − n i for 1 ≤ i ≤ L. From (28), we replace ˜k i M Γ i by Γσx − n i Γ i − ǫ i Γ i in both sides of (27) <strong>and</strong> have Thus, using (9) <strong>and</strong> (17) we have |µ i Γ i − µ 1 Γ 1 − (ǫ i Γ i − ǫ 1 Γ 1 )| ≤ |ǫ i Γ i − ǫ 1 Γ 1 |. |µ i Γ i − µ 1 Γ 1 | ≤ 2|ǫ i Γ i − ǫ 1 Γ 1 | ≤ 1 + 2τ M (Γ 1 + Γ i ) < 1. (29) Since µ i , Γ i , µ 1 , <strong>and</strong> Γ 1 are all integers, (29) implies µ i Γ i = µ 1 Γ 1 , i = 2,3,...,L. (30)
8 Since Γ i <strong>and</strong> Γ 1 are co-prime, (30) implies µ 1 = m i Γ i <strong>and</strong> µ i = m i Γ 1 , i.e., ¯n 1 = n 1 + m i Γ i <strong>and</strong> ¯n i = n i + m i Γ 1 (31) for some integers m i with |m i | < min(γ i ,γ 1 ). Replacing (31) into (27), we find that ∣ n iΓ i + ˜k i Γ i M − n 1Γ 1 − ˜k ∣ 1 Γ 1 ∣∣∣∣¯n M ∣ = i Γ i + ˜k i Γ i M − ¯n 1Γ 1 − ˜k 1 Γ 1 M ∣ , (32) which means (n 1 ,n i ) ∈ S i for i = 2,3,...,L. This proves n 1 ∈ S. We next show S = {n 1 }. Property (31) also implies S i = {(n 1 + m i Γ i , n i + m i Γ 1 ) : for some integers m i with |m i | < min(γ i ,γ 1 )}. (33) If ¯n 1 ∈ S, then ¯n 1 ∈ S i,1 for i = 2,3,...,L, <strong>and</strong> therefore, from the definition of S i,1 in (14) <strong>and</strong> (33), we have ¯n 1 − n 1 = m i Γ i for some integer m i with |m i | < min(γ i ,γ 1 ) for i = 2,3,...,L. This implies that ¯n 1 − n 1 divides all Γ i for i = 2,3,...,L <strong>and</strong> therefore from (12), ¯n 1 − n 1 is a multiple of γ 1 . Since 0 ≤ ¯n 1 ,n 1 ≤ γ 1 − 1, we conclude ¯n 1 − n 1 = 0. This proves that S = {n 1 }. In the meantime, ¯n 1 = n 1 implies m i = 0 in (33), i.e., ¯n i = n i for i = 2,3,...,L. Hence, <strong>Theorem</strong> 1 is proved. As a remark, despite [4] only considers the case when τ = 0, the proof in [4] has errors. REFERENCES [1] O. Goldreich, D. Ron, <strong>and</strong> M. Sudan, “<strong>Chinese</strong> remaindering with errors,” IEEE Trans. Information Theory, vol.46, pp.1330-1338, July 2000. [2] V. Guruswami, A. Sahai, <strong>and</strong> M. Sudan, “’Soft-decision’ decoding of <strong>Chinese</strong> remainder codes,” Proc. 41st IEEE Symp. on Found. of Comp. Sci., pp. 159-168, Redondo Beach, California, 2000. [3] I. E. Shparlinski <strong>and</strong> R. Steinfeld, “Noisy <strong>Chinese</strong> <strong>Remainder</strong>ing in the Lee Norm,” preprint. [4] G. Wang, X.-G. Xia, V. C. Chen, <strong>and</strong> R. L. Fiedler, “Detection, location, <strong>and</strong> imaging of fast moving targets using multifrequency antenna array SAR,” IEEE Trans. Aerospace <strong>and</strong> Electronic Systems, vol. 40, pp. 345-355, Jan. 2004. [5] X.-G. Xia, “Estimation of multiple frequencies in undersampled complex valued waveforms,” IEEE Trans. Signal Processing, vol. 47, pp. 3417-3419, Dec. 1999. [6] X.-G. Xia, “An efficient frequency determination algorithm from multiple undersampled waveforms,” IEEE Signal Processing Lett., vol. 7, pp. 34-37, Feb. 2000. [7] X.-G. Xia <strong>and</strong> K. Liu, “A generalized <strong>Chinese</strong> remainder theorem for residue sets with errors <strong>and</strong> its application in frequency determination from multiple sensors with low sampling rates,” IEEE Signal Processing Lett., vol. 12, pp.768-771, Nov. 2005. [8] W. Xu, E. C. Chang, L. K. Kwoh, H. Lim, <strong>and</strong> W. C. A. Heng, “<strong>Phase</strong> unwrapping of SAR interferogram with multifrequency or multi-baseline,” Proc. IGASS’94, pp. 730-732, 1994. [9] J. H. McClellan <strong>and</strong> C. M. Rader, Number Theory in Digital Signal Processing, Prentice-Hall, Englewood Cliffs, N. J., 1979. [10] C. Ding, D. Pei, <strong>and</strong> A. Salomaa, <strong>Chinese</strong> <strong>Remainder</strong> <strong>Theorem</strong>: Applications in Computing, Coding, Cryptography, World Scientific Pub., 1999.