Skip to main content
Log in

Insights on coding gain and its properties for principal component filter banks

  • Published:
Sādhanā Aims and scope Submit manuscript

Abstract

Principal Component Filter Bank (PCFB) is considered optimal in terms of coding gain for specific conditions. P P Vaidyanathan stated that coding gain does not necessarily always increase with the increase in the number of bands. However, very few attempts are made in the literature to go beyond the confines of work done by P P Vaidyanathan. We present the analytic proofs for the monotonicity of specific shapes of PSDs. This papers also derives properties of coding gain of PCFBs, which brings the new insights on the coding gain of Principal Component Filter Banks.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Figure 1
Figure 2
Figure 3
Figure 4
Figure 5
Figure 6

Similar content being viewed by others

References

  1. Tsatsanis M K and Giannakis G B 1995 Principal component filter banks for optimal multiresolution analysis. IEEE Trans. Signal Process. 43(8): 1766–1777

    Article  Google Scholar 

  2. Vaidyanathan P P and Akkarakaran S 2001 A review of the theory and applications of optimal subband and transform coders. Appl. Comput. Harmon. Anal. 10(3): 254–289

    Article  MathSciNet  MATH  Google Scholar 

  3. Akkarakaran S and Vaidyanathan P P 2001 Filterbank optimization with convex objectives and the optimality of principal component forms. IEEE Trans. Signal Process. 49(1): 100–114

    Article  MathSciNet  MATH  Google Scholar 

  4. Vaidyanathan P P 1998 Theory of optimal orthonormal subband coders. IEEE Trans. Signal Process. 46(6): 1528–1543

    Article  Google Scholar 

  5. Xuan B and Bamberger R I 1998 FIR principal component filter banks. IEEE Trans. Signal Process. 46(4): 930–940

    Article  Google Scholar 

  6. Tkacenko A 2004 Optimization algorithms for realizable signal-adapted filter banks. California Institute of Technology, Pasadena, United States.

    Google Scholar 

  7. Vouras P G and Tran T D 2006 Design of FIR paraunitary approximations to principal component filter banks. In: 2006 40th Annual Conference on Information Sciences and Systems, pp. 1372–1377

  8. Vouras P G and Tran T D 2005 Principal component filter bank for band partitioned sidelobe cancellation. In: IEEE International Radar Conference, 2005, pp. 691–696

  9. Jhawar A, Ginde P, Patwardhan P and Gadre V M 2010 Coding gain optimized finite impulse response (FIR) paraunitary (PU) filter banks. In: 2010 National Conference on Communications (NCC), pp. 1–5

  10. Redif S, McWhirter J G and Weiss S 2011 Design of FIR paraunitary filter banks for subband coding using a polynomial eigenvalue decomposition. IEEE Trans. Signal Process., 59(11): 5253–5264

    Article  MathSciNet  MATH  Google Scholar 

  11. Fatimah B and Joshi S 2015 An optimal biorthogonal M-channel signal matched FIR filter bank. In: 2015 International Conference on Computational Science and Computational Intelligence (CSCI), pp. 443–447

  12. Vaidyanathan P P 1993 Multirate Systems and Filter Banks. Engelwood Cliffs, NJ: Prentice Hall PTR.

    MATH  Google Scholar 

  13. Chaphekar P 2020 New Results and Design Techniques for Principal Component Filter Banks. PhD thesis, Indian Institute of Technology, Bombay, India.

Download references

Author information

Authors and Affiliations

Authors

Corresponding author

Correspondence to Ashish V Vanmali.

Appendix: Subband variance of triangular PSD

Appendix: Subband variance of triangular PSD

Consider a triangular PSD with the subbands designed for PCFB. For M subbands, the frequency support of each band will be \(\frac{\pi }{M}\). For \(M=5\), these subbands will appear as shown in figure 7. The first subband will be from \(\omega _4\) to \(\omega _5\). The second subband will be from \(\omega _3\) to \(\omega _4\) and \(\omega _5\) to \(\omega _6\). The third subband will be from \(\omega _2\) to \(\omega _3\) and \(\omega _6\) to \(\omega _7\), and so no. Then,

\( \text {Subband variance} = \frac{1}{\pi } \times \text {Area under the PSD for subband}. \)

Thus,

$$\begin{aligned} \sigma _{1}^{2}&= \frac{1}{\pi } ( \text {Area of triangle} + \text {Area of rectangle} ) \nonumber \\&= \frac{1}{\pi } \left[ \frac{1}{2} \cdot \frac{\pi }{M} \cdot \frac{a}{M} + \frac{\pi }{M} \cdot \frac{a(M-1)}{M} \right] \nonumber \\&= a \left[ \frac{1+2(M-1)}{2M^2} \right] . \end{aligned}$$
(79)
Figure 7
figure 7

Subbands of triangular PSD.

The area of the second subband can be computed by finding the area of parts numbered 1 and 2 combined and then subtracting the area of the first subband. Thus,

$$\begin{aligned} \sigma _{2}^{2}&= \frac{1}{\pi } \left[ \frac{1}{2} \cdot \frac{2\pi }{M} \cdot \frac{2a}{M} + \frac{2\pi }{M} \cdot \left( a - \frac{2a}{M} \right) \right] - a\left[ \frac{1+2(M-1)}{2M^2}\right] \nonumber \\&= a \left[ \frac{1+2(M-2)}{2M^2} \right] . \end{aligned}$$
(80)

The area of subsequent subbands can be found in a similar manner. In general,

$$\begin{aligned} \sigma _{M-k}^{2}&= a \left[ \frac{1+2k}{2M^2} \right] \text {where } k=0,1,\ldots ,M-1. \end{aligned}$$
(81)

Rights and permissions

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Chaphekar, P., Bhatia, A., Bhatnagar, S. et al. Insights on coding gain and its properties for principal component filter banks. Sādhanā 48, 76 (2023). https://doi.org/10.1007/s12046-023-02142-y

Download citation

  • Received:

  • Revised:

  • Accepted:

  • Published:

  • DOI: https://doi.org/10.1007/s12046-023-02142-y

Keywords

Navigation