Optimal maximally decimated M-channel mirrored paraunitary linear phase FIR filter bank design via norm relaxed sequential quadratic programming

Qing Liu; Bingo Wing-Kuen Ling; Qingyun Dai; Qing Miao; Caixia Liu

doi:10.3934/jimo.2020055

Article Contents

2021, Volume 17, Issue 4: 1993-2011. Doi: 10.3934/jimo.2020055

This issue Previous Article A better dominance relation and heuristics for Two-Machine No-Wait Flowshops with Maximum Lateness Performance Measure Next Article Local search algorithm for the squared metric

$ k $

-facility location problem with linear penalties

Optimal maximally decimated M-channel mirrored paraunitary linear phase FIR filter bank design via norm relaxed sequential quadratic programming

Faculty of Information Engineering, Guangdong University of Technology, Guangzhou, 510006, China

^* Corresponding author: Bingo Wing-Kuen Ling
^* Corresponding author: Bingo Wing-Kuen Ling

Received: July 31, 2019

Revised: October 31, 2019

Early access: March 2020

Published: July 2021

Abstract Full Text(HTML) Figure(2) / Table(3) Related Papers Cited by

Abstract

It is worth noting that the conventional maximally decimated M-channel mirrored paraunitary linear phase finite impulse response condition is defined in the frequency domain. As the frequency domain is a continuous set, it is expressed as a matrix functional (a continuous function of the frequency) equation. On the other hand, this paper expresses the condition as a finite number of discrete (a set of functions of the sampled frequencies) equations. Besides, this paper proposes to sample the magnitude responses of the filters with the total number of the sampled frequencies being more than the filter lengths. Hence, the frequency selectivities of the filters can be controlled more effectively. This filter bank design problem is formulated as an optimization problem in such a way that the total mirrored paraunitary linear phase error is minimized subject to the specifications on the magnitude responses of the filters at these sampling frequencies. However, this optimization problem is highly nonconvex. To address this difficulty, a norm relaxed sequential quadratic programming approach is applied for finding its local optimal solution. By iterating the above procedures using different initial conditions, a near global optimal solution is obtained. Computer numerical simulation results show that our proposed design outperforms the existing designs.

Keywords:

Mathematics Subject Classification: Primary: 58F15, 58F17; Secondary: 53C35.

Citation:

Full Text(HTML)

Figure 1. $ 20{\log _{10}}\left( {{{\left| {{H_m}\left( \omega \right)} \right|} \over {\sqrt M }}} \right) $ for $ m = 0, \dotsc, 3 $ of the analysis filters in decibels designed by our proposed method as well as those designed by the methods discussed in both [14] and [30]

Download: Full-size image PowerPoint slide

Figure 2. (a) $ 10{\log _{10}}\left( {\left| {\left| {{1 \over M}\sum\limits_{m = 0}^{M - 1} {{H_m}\left( \omega \right){{\tilde H}_m}\left( \omega \right)} } \right| - 1} \right|} \right) $ and (b) $ 10{\log _{10}}\left( {\left| {{1 \over M}\sum\limits_{k = 1}^{M - 1} {\sum\limits_{m = 0}^{M - 1} {{H_m}\left( {\omega - {{2\pi k} \over M}} \right){{\tilde H}_m}\left( \omega \right)} } } \right|} \right) $ of the filter banks in decibels designed by our proposed method as well as those designed by the methods discussed in both [14] and [30]

Download: Full-size image PowerPoint slide

Table 1. $ \mathop {\max }\limits_{\omega \in B_m^p \cup B_m^s} 20{\log _{10}}\left( {\big| {\left| {{H_m}\left( \omega \right)} \right| - \left| {{D_m}\left( \omega \right)} \right|} \big|} \right) $ for $ m = 0, \dotsc, 3 $ of the analysis filters in decibels designed by our proposed method as well as those designed by the methods discussed in both [14] and [30]

	The maximum ripple magnitude of the first analysis filter in decibel	The maximum ripple magnitude of the second analysis filter in decibel	The maximum ripple magnitude of the third analysis filter in decibel	The maximum ripple magnitude of the fourth analysis filter in decibel
Method discussed in [14]	-0.3251dB	-11.3525dB	-11.3525dB	-0.3251dB
Method discussed in [30]	-0.7366dB	-13.0137dB	-12.9932dB	-0.7372dB
Our proposed method	-6.1628dB	-6.1930dB	-6.1930dB	-6.1628dB

| Show Table

DownLoad: CSV

Table 2. $ {\log_{10}}\left( {err_{para}\left( l \right)} \right) $ for $ l = 0, \dotsc, L-1 $ of the filter banks in decibels designed by our proposed method as well as those designed by the methods discussed in both [14] and [30]

	Method discussed in [22]	Method discussed in [16]	Our proposed method
$ {\log_{10}}\left( {err_{para}\left( 0 \right)} \right) $	-66.7193dB	-141.8743dB	1.1415dB
$ {\log_{10}}\left( {err_{para}\left( l \right)} \right) $	-141.4721dB	-140.8206dB	-0.8275dB

| Show Table

DownLoad: CSV

Table 3. $ \mathop {\max }\limits_\omega 10{\log _{10}}\left( {\left| {\left| {{1 \over M}\sum\limits_{m = 0}^{M - 1} {{H_m}\left( \omega \right){{\tilde H}_m}\left( \omega \right)} } \right| - 1} \right|} \right) $ and $ \mathop {\max }\limits_\omega 10{\log _{10}}\left( {\left| {{1 \over M}\sum\limits_{k = 1}^{M - 1} {\sum\limits_{m = 0}^{M - 1} {{H_m}\left( {\omega - {{2\pi k} \over M}} \right){{\tilde H}_m}\left( \omega \right)} } } \right|} \right) $ of the filter banks in decibels designed by our proposed method as well as those designed by the methods discussed in both [14] and [30]

	Method discussed in [14]	Method discussed in [30]	Our proposed method
$ \mathop {\max }\limits_\omega 10{\log _{10}}( \| \| {{1 \over M}\sum\limits_{m = 0}^{M - 1} {{H_m}(\omega){{\tilde H}_m}(\omega)} } \| - 1 \| ) $	-72.7399dB	-142.8249dB	-2.9505dB
$ \mathop {\max }\limits_\omega 10{\log _{10}}( {\| {{1 \over M}\sum\limits_{k = 1}^{M - 1} {\sum\limits_{m = 0}^{M - 1} {{H_m}( {\omega - {{2\pi k} \over M}} ){{\tilde H}_m}( \omega )} } } \|}) $	-146.7204dB	-145.1184dB	-6.0642dB

| Show Table

DownLoad: CSV

Related Papers

Cited by

References

[1]	Y.-J. Chen, S. Oraintara and K. S. Amaratunga, Dyadic-based factorizations for regular paraunitary filterbanks and $M$-band orthogonal wavelets with structural vanishing moments, IEEE Transactions on Signal Processing, 53 (2005), 193-207. doi: 10.1109/TSP.2004.838962.
[2]	M. T. de Gouvêa and D. Odloak, A new treatment of inconsistent quadratic programs in a sqp-based algorithm, Computers & Chemical Engineering, 22 (1998), 1623-1651.
[3]	Y.-T. Fong and C.-W. Kok, Correction to "Iterative least squares design of DC-leakage free paraunitary cosine modulated filter banks'', IEEE Transactions on Circuits and Systems II: Analog and Digital Signal Processing, 50 (2003), 238-243. doi: 10.1109/TCSII.2007.895968.
[4]	X. Q. Gao, T. Q. Nguyen and G. Strang, Theory and lattice structure of complex paraunitary filterbanks with filters of (hermitian-) symmetry/antisymmetry properties, IEEE Transactions on Signal Processing, 49 (2001), 1028-1043. doi: 10.1109/78.917806.
[5]	X. Q. Gao, T. Q. Nguyen and G. Strang, On factorization of $M$-channel paraunitary filterbanks, IEEE Transactions on Signal Processing, 49 (2001), 1433-1446. doi: 10.1109/78.928696.
[6]	L. Gan and K.-K. Ma, A simplified lattice factorization for linear-phase paraunitary filter banks with pairwise mirror image frequency responses, IEEE Transactions on Circuits and Systems II: Express Briefs, 51 (2004), 3-7. doi: 10.1109/TCSII.2003.821515.
[7]	N. Holighaus, Z. Prŭša and P. L. Søndergaard, Reassignment and synchrosqueezing for general time-frequency filter banks, subsampling and processing, Signal Processing, 125 (2016), 1-8. doi: 10.1016/j.sigpro.2016.01.007.
[8]	M. Ikehara, T. Nagai and T. Q. Nguyen, Time-domain design and lattice structure of FIR paraunitary filter banks with linear phase, Signal Processing, 80 (2000), 333-342. doi: 10.1016/S0165-1684(99)00131-0.
[9]	M. Ikehara and T. Q. Nguyen, Time-domain design of linear-phase pr filter banks, 1997 IEEE International Conference on Acoustics, Speech, and Signal Processing, 3 (1997), 2077-2080.
[10]	J.-Z. Jiang, F. Zhou, S. Ouyang and G. S. Liao, Efficient design of high-complexity interleaved DFT modulated filter bank, Signal Processing, 94 (2014), 130-137. doi: 10.1016/j.sigpro.2013.06.006.
[11]	J.-B. Jian, Q.-J. Xu and D.-L. Han, A strongly convergent norm-relaxed method of strongly sub-feasible direction for optimization with nonlinear equality and inequality constraints, Applied Mathematics and Computation, 182 (2006), 854-870. doi: 10.1016/j.amc.2006.04.049.
[12]	J.-B. Jian, X.-Y. Ke and W.-X. Cheng, A superlinearly convergent norm-relaxed sqp method of strongly sub-feasible directions for constrained optimization without strict complementarity, Applied Mathematics and Computation, 214 (2009), 632-644. doi: 10.1016/j.amc.2009.04.022.
[13]	C. W. Kok, T. Nagai, M. Ikehara and T. Q. Nguyen, Lattice structures parameterization of linear phase paraunitary matrices with pairwise mirror-image symmetry in the frequency domain with an odd number of rows, IEEE Transactions on Circuits and Systems II: Analog and Digital Signal Processing, 48 (2001), 633-636. doi: 10.1109/82.943336.
[14]	Y.-P. Lin and P. Vaidyanathan, Linear phase cosine modulated maximally decimated filter banks with perfect reconstruction, IEEE Transactions on Signal Processing, 43 (1995), 2525-2539.
[15]	C. Liu, B. W. Ling, C. Y. Ho and Q. Dai, Finite number of necessary and sufficient discrete condition in frequency domain for maximally decimated m-channel mirrored paraunitary linear phase fir filter bank, 2014 IEEE International Conference on Consumer Electronics-China, (2014), 1–5.
[16]	B. W.-K. Ling, N. Tian, C. Y.-F. Ho, W.-C. Siu, K.-L. Teo and Q. Y. Dai, Maximally decimated paraunitary linear phase FIR filter bank design via iterative SVD approach, IEEE Transactions on Signal Processing, 63 (2015), 466-481. doi: 10.1109/TSP.2014.2371779.
[17]	T. Q. Nguyen, A quadratic-constrained least-squares approach to the design of digital filter banks, 1992 IEEE International Symposium on Circuits and Systems, 3 (1992), 1344-1347. doi: 10.1109/ISCAS.1992.230255.
[18]	T. Q. Nguyen, A. K. Soman and P. Vaidyanathan, A quadratic-constrained least-squares approach to linear phase orthonormal filter bank design, 1993 IEEE International Symposium on Circuits and Systems, (1993), 383–386.
[19]	S. Oraintara, T. D. Tran, P. N. Heller and T. Q. Nguyen, Lattice structure for regular paraunitary linear-phase filterbanks and $M$-band orthogonal symmetric wavelets, IEEE Transactions on Signal Processing, 49 (2001), 2659-2672. doi: 10.1109/78.960413.
[20]	S. Patel, R. Dhuli and B. Lall, Design and analysis of matrix wiener synthesis filter for multirate filter bank, Signal Processing, 102 (2014), 256-264. doi: 10.1016/j.sigpro.2014.03.021.
[21]	M. Sangnier, J. Gauthier and A. Rakotomamonjy, Filter bank learning for signal classification, Signal Processing, 113 (2015), 124-137. doi: 10.1016/j.sigpro.2014.12.028.
[22]	A. K. Soman, P. P. Vaidyanathan and T. Q. Nguyen, Linear phase paraunitary filter banks: Theory, factorizations and designs, IEEE Transactions on Signal Processing, 41 (1993), 3480-3496. doi: 10.1109/78.258087.
[23]	A. K. Soman and P. P. Vaidyanathan, A complete factorization of paraunitary matrices with pairwise mirror-image symmetry in the frequency domain, IEEE Transactions on Signal Processing, 43 (1995), 1002-1004. doi: 10.1109/78.376855.
[24]	C. G. Shen, W. J. Xue and X. D. Chen, Global convergence of a robust filter SQP algorithm, European Journal of Operational Research, 206 (2010), 34-45. doi: 10.1016/j.ejor.2010.02.031.
[25]	T. D. Tran and T. Q. Nguyen, On m-channel linear phase fir filter banks and application in image compression, IEEE Transactions on Signal Processing, 45 (1997), 2175-2187.
[26]	T. D. Tran, M. Ikehara and T. Q. Nguyen, Linear phase paraunitary filter bank with filters of different lengths and its application in image compression,, IEEE Transactions on Signal Processing, 47 (1999), 2730-2744. doi: 10.1109/78.790655.
[27]	T. D. Tran, R. L. De Queiroz and T. Q. Nguyen, Linear-phase perfect reconstruction filter bank: Lattice structure, design, and application in image coding, IEEE Transactions on Signal Processing, 48 (2000), 133-147.
[28]	P. G. Vouras and T. D. Tran, Factorization of paraunitary polyphase matrices using subspace projections, 2008 42nd Asilomar Conference on Signals, Systems and Computers, (2008), 602–605. doi: 10.1109/ACSSC.2008.5074476.
[29]	Z. M. Xu and A. Makur, On the arbitrary-length $M$-channel linear phase perfect reconstruction filter banks, IEEE Transactions on Signal Processing, 57 (2009), 4118-4123. doi: 10.1109/TSP.2009.2024026.
[30]	W. J. Xue, C. G. Shen and D. G. Pu, A penalty-function-free line search sqp method for nonlinear programming, Journal of Computational and Applied Mathematics, 228 (2009), 313-325. doi: 10.1016/j.cam.2008.09.031.