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
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
3.2. NONLINEAR MODELING OF TIME SERIES 39<br />
where Linear, Quadratic, and Cubic denote the matrices <strong>of</strong> linear, quadratic,<br />
and cubic filter coefficients respectively:<br />
and<br />
⎢<br />
Quadratic = ⎢<br />
⎣<br />
⎡<br />
⎢<br />
Cubic = ⎢<br />
⎣<br />
⎡<br />
⎢<br />
Linear = ⎢<br />
⎣<br />
⎡<br />
xl+1 xl+2 · · · xl+p<br />
.<br />
. ..<br />
xm xm · · · xm<br />
.<br />
.<br />
xn−1 xn−2 · · · xn−p<br />
x 2 l+1 2xl+1xl+2 2xl+1xl+3 · · · x 2 l+q<br />
.<br />
. ..<br />
x 2 m 2xmxm 2xmxm · · · x 2 m<br />
.<br />
.<br />
.<br />
.<br />
.<br />
.<br />
.<br />
.<br />
.<br />
. ..<br />
x 2 n−1 2xn−1xn−2 2xn−1xn−3 · · · x 2 n−q<br />
⎤<br />
⎥ , (3.26)<br />
⎥<br />
⎦<br />
.<br />
.<br />
⎤<br />
⎥ , (3.27)<br />
⎥<br />
⎦<br />
x 3 l+1 · · · 3x 2 l+1xl+2 3x 2 l+1xl+3 · · · 6xl+1xl+2xl+3 · · · x 3 l+r<br />
.<br />
.<br />
.<br />
.<br />
x 3 m · · · 3x 2 mxm 3x 2 mxm · · · 6xmxmxm · · · x 3 m<br />
.<br />
.<br />
.<br />
.<br />
x 3 n−1 · · · 3x 2 n−1xn−2 3x 2 n−1xn−3 · · · 6xn−1xn−2xn−3 · · · x 3 n−r<br />
.<br />
.<br />
.<br />
.<br />
.<br />
. ..<br />
.<br />
.<br />
⎤<br />
⎥ .<br />
⎥<br />
⎦<br />
(3.28)<br />
The unknown vector m contains linear and nonlinear prediction coefficients orga-<br />
nized in lexicographic form in the dimension <strong>of</strong> P + Q(Q+3)<br />
2 − Q + R2 + R! ) × 1.<br />
(R−3)!3!