Download presentation
Presentation is loading. Please wait.
1
Spline Interpretation ABC Introduction and outline Based mostly on Wikipedia
2
Spline Interpretation Spline is a flat flexible strip of thin narrow piece of wood, metal, or plastic used in drawing curved. Mathematically, spline means curve fitting mostly using certain piecewise low degree polynomials. The main reason to use low degree polynomial is to avoid Runge’s phenomenon.
3
Runge’s Phenomenon
4
Runge’s Phenomenon Red: data; blue: 5 th order; green: 9 th order
5
Runge’s Phenomenon
6
Definition of Spline
7
Linear Spline
8
Quadratic Spline
10
Cubic Spline
13
Hermite Spline
15
Hermite spline
18
In general, cubic spline does not guarantee monotonicity even for monotonic data. Hermite monotonic spline, however, guarantees monotonicity between data points. The advantage is the elimination of overshoots. The prices to pay are these: – we no longer have the smoothness of continuity in curvature and –we have to inject another subjective idea of monotonic and/or variations.
19
Comparisons between Cubic Natural and Hermite Splines An Example
20
Data
21
Envelopes of Maxima
22
Envelopes of Minima
23
Mean of Max and Min Envelopes
24
Difference between Mean of Max and Min Envelopes
25
First Proto-IMF
26
Observations Hermite monotonic spline seems to have better performance with limited variations. The end effect is especially severe for natural spline, a problem we have to resolve. In order to be fully adaptive with minimum subjective intervention, we selected cubic natural spline for EMD with only one assumption: smoothness.
27
B-Spline B-spline is Basis-spline. Something used extensively by Computer Aided Design (CAD), but we have found limited use of it.
28
Bézier Curve
30
Cubic Bezier Bases
31
Bézier Curve
32
An example
33
Rational Bézier Curve
34
B-Spline A B-spline is a generalization of the Bézier Curve; A B-spline with no internal knots is a Bézier Curve. It is generated through Cox-de Boor recursion formula. A further generalization is NURBS: NonUniform Rational B-Spline. See for example, David F. Rogers: An Introduction to NURBS. Morgan Kaufmann, San Francisco, 2001
35
B-Spline
36
Properties of B-Spline A B-spline could have degree n; control points, p; and knots, m, each independent to others, but m ≥ p. It is much more flexible. For n=3, p=10, and m=14, we have B-spline at left and Bezier curve at right.
37
Properties of B-Spline In general, the lower the degree, the closer the curve follows the control points. For the same numbers of control points and knots, but different degrees, the spline results are very different: left, n=7; center, n=5; right, n=3.
38
Properties of B-Spline B-spline is a piecewise curve. We must have m=n+p+1. Clamped spline has to pass the beginning and end points. Strong convex hull property Variation diminishing property Affine invariance But B-Spline does not pass through the control points!
39
References Qiuhui Chen, Norden Huang, Sherman Riemenschneider and Yuesheng Xu, 2006: A B-spline approach for empirical mode decompositions. Advances in Computational Mathematics 24: 171–195
40
Variation Diminishing Property: no straight line intersects a B-spline curve more times than it intersects the curve's control polygon.
41
Summary The differences between different splines seems to be huge. Based on principles of least interferences and maximum smoothness, we selected natural spline as the base for EMD. Different splines (except B-spline) should not change the scales of the IMFs, but it would change the shape and energy levels. EMD is unique with respect to specific spline fitting and parameters selected.
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.