Overview of Body Sway Data Analysis

1
Overview of Body Sway Data Analysis

• Robert Harrington
• Winter 2003

2
Goals

• Ultimate goal to determine likelihood of a
  persons falling by analyzing small motion while
  standing

FOR MORE INFO...
Prof. David Garelick Northeastern
University 617.373.2936
3
Body Position versus Time
4
Fourier Transform
5
Fourier Transform of Data (cont.)
6
Fourier Transform of Data (cont.)
7
Fourier Transform of Data (cont.)
8
Fourier Transform of Data (cont.)
9
Fourier Transform of Data (cont.)
10
Difficulties with Fourier transform

• No statistically meaningful differences
  between fallers and non-fallers
• Fixed range of frequencies
• 1/(2?t) to 1/(2?t)
• Does not directly include exponential
  damping

11
Proposed Model

• Body can be modeled as a simple spring with the
  restoring force proportional to the displacement
  at an earlier time
• F - k x ( t - ? )
• This model gives solutions which have the form
  a e-?t sin(?o t)

12
Laplace Transform

• Allows determination of exponential damping term
• Problem Discrete Laplace Transform very
  difficult

13
Laplace Transform graph
14
Laplace Transform (cont.)
15
Laplace Transform problem

• Analytical solution with non-infinite
  integration limits gives extra term
• Extra term depends on limits AND position in
  frequency space
• Creates peaks at spurious frequencies
• As limit goes to infinity, peaks converge to one
  at the correct frequency

16
Laplace Transform problem
17
Laplace Transform problem
18
Current Status

• Analysis of Fourier peak-widths to determine
  exponential damping term
• Fitting curves to function, using model to
  predict ? corresponding to know Fourier peaks
  (parameter ?)