R. Andries
10-10-1995, 06:35 PM
Dear biomch-L readers,

I'm new to the biomch network. I'm working on a Ph. D. on knee kinematics in
vivo using positional data of markers (measured with Vicon) as input.
I have two problems in calculating knee kinematics:

1. Until now I calculated the rotation matrix using the least squares
method described by Veldpaus e.a. (J. Biomech 21: 45-54). In several
publications (Challis 1995 J. Biomech 28 733-737; Soderkvist and Wedin
1993 J. Biomech. 26: 1473-1477; Woltring in several publications; ...)
the Singular-Value-Decomposition (SVD) theory is proposed to estimate
R, in stead of the polar decomposition theorem proposed by Veldpaus.
Do both procedures give different results? And suppose SVD is better,
(as sometimes is mentioned), can somebody give me some practical
information that can help to program the SVD-calculation in
Borland Pascal (Dos version).

2. When I calculate the instantaneous screw axes parameters I need to
estimate the angular velocity vector omega. I use the formula
published in Woltring e.a. (1994, J. Biomech. 27: 1415-1432) where
the velocity vector is approximated from (R+R'_ - R_R'+)*(1/4T)

where R+ and R_ are the attitude matrix of the next and
previous sampling times and where T is the sampling

I'd like to know if there is a better way of calculating the velocity
vector using the velocity vector of each marker as input AND
using a least-squares technique.

Thanks in advance for your reply,

Kaat Desloovere
Faculty of Physical Education and Physical Therapy K.U.Leuven
Tervuursevest 101
3001 Leuven (Heverlee)
tel -- 32 16 32 90 74
fax -- 32 16 32 91 96
email kaat.desloovere@flok.kuleuven.ac.be