Comparison of Daubechies wavelets for Hurst parameter estimation


Ciftlikli C., GEZER A.

TURKISH JOURNAL OF ELECTRICAL ENGINEERING AND COMPUTER SCIENCES, cilt.18, ss.117-128, 2010 (SCI İndekslerine Giren Dergi) identifier

  • Cilt numarası: 18 Konu: 1
  • Basım Tarihi: 2010
  • Doi Numarası: 10.3906/elk-0905-47
  • Dergi Adı: TURKISH JOURNAL OF ELECTRICAL ENGINEERING AND COMPUTER SCIENCES
  • Sayfa Sayıları: ss.117-128

Özet

Time scale dependence on the working nature of wavelet analysis makes it a valuable tool for Hurst parameter estimation Similar to other wavelet-based signal processing applications, the selection of a particular wavelet type and vanishing moment m wavelet based Hurst estimation is a challenging problem In this paper, we investigate the best Daubechies wavelet in wavelet based Hurst estimation for an exact self similar process, fractional Gaussian noise and how Daubechies vanishing moment affects the Hurst estimation accuracy Daubechies wavelets are preferred in analysis because increasing vanishing moment does not cause excessive increase of time support of Daubechies wavelets Thus, limited time support of wavelets reduces the border effects Results show that Daubechies wavelets with one vanishing moment (Daubechies 1) gives the best estimation result for short range dependent fractional Gaussian. noise Daubechies 2 is the best preference for long range dependent fractional Gaussian noise

Time scale dependence on the working nature of wavelet analysis makes it a valuable tool for Hurst

parameter estimation. Similar to other wavelet-based signal processing applications, the selection of a

particular wavelet type and vanishing moment in wavelet based Hurst estimation is a challenging problem.

In this paper, we investigate the best Daubechies wavelet in wavelet based Hurst estimation for an exact self

similar process, fractional Gaussian noise and how Daubechies vanishingmoment affects the Hurst estimation

accuracy. Daubechies wavelets are preferred in analysis because increasing vanishing moment does not cause

excessive increase of time support of Daubechies wavelets. Thus, limited time support of wavelets reduces

the border effects. Results show that Daubechies wavelets with one vanishing moment (Daubechies 1) gives

the best estimation result for short range dependent fractional Gaussian noise. Daubechies 2 is the best

preference for long range dependent fractional Gaussian noise.