18.08.2013 Views

Soner Bekleric Title of Thesis: Nonlinear Prediction via Volterra Ser

Soner Bekleric Title of Thesis: Nonlinear Prediction via Volterra Ser

Soner Bekleric Title of Thesis: Nonlinear Prediction via Volterra Ser

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

Chapter 6<br />

Conclusions and Future Directions<br />

This thesis presented a method <strong>of</strong> signal modeling based on <strong>Volterra</strong> series. Specif-<br />

ically, I have developed an autoregressive method based on a nonlinear prediction<br />

algorithm. The prediction coefficients <strong>of</strong> the first-, second-, and third-order <strong>Volterra</strong><br />

model are obtained with a least squares solution. I also bring a new approach to<br />

solving two fundamental problems in seismic exploration: modeling <strong>of</strong> complex<br />

waveforms in the f − x domain and adaptive subtraction <strong>of</strong> multiples.<br />

Primary issues and available methods in modeling time series are examined<br />

in Chapters 1 and 2. First, the nonlinear system is introduced as a higher-order<br />

extension <strong>of</strong> the usual linear convolution method. Former linear prediction methods<br />

are also investigated to obtain prediction coefficients. Comparisons between a first-<br />

order <strong>Volterra</strong> series and Yule-Walker equations and Burg’s algorithm have shown<br />

that linear prediction methods model the data quite similarly. On the contrary,<br />

all linear methods can not properly model complex data sets (waveforms) unless a<br />

large number <strong>of</strong> coefficients is used.<br />

In Chapter 3 a <strong>Volterra</strong> series and its properties are analyzed. The <strong>Volterra</strong><br />

91

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!