To drive the biomechanical man with real data, the local coordinate systems of the IMUs were mapped to the local coordinate systems of the body segments through the calibration procedure described below and pictured in (Figure 4.11).
In the calibration procedure if soft tissue artefacts are minimised, a constant rotation „maps‟ the IMUs to the different body segments and an accurate measurement of limb orientation is possible. The constant rotation is defined below by a 3-by-3 rotation matrix (RILBL), which
transforms the measurements from the IMU local (IL) to the body segment local (BL) coordinate systems. The calibration process requires five seconds of stationary data from the IMUs attached to the athlete. During calibration the athlete is precisely supported in the reference position by the 3D anthropometric frame (Figure 4.11) and the following four steps are completed:
1. The local orientations of the body segments with respect to the 3D anthropometric frame (F), and called (RBLF). Measurements of bony landmarks from the
anthropometric frame are used.
2. The mean orientation of the IMUs (RILG), are calculated from the IMUs
accelerometer and magnetometer channels. The static orientation calculation described in Chapter 2 on page 30 is used.
3. The orientation of the anthropometric frame with respect to the IMU global coordinate system (RFG), is required. It is defined as the rotation about the vertical
axis that brings the X-axis of the frame in line with magnetic north. This can be determined by using a compass to measure the frame X-axis heading.
4. The calibration matrix for each body segment can now be determined (Equation 4.10).
Equation 4.10
The procedures are now in place for FMC data from skiing to drive animations of the biomechanical man. In the future, the captured motion will be reconstructed by transforming the motion of each body segment into the global coordinate system. The body segment defined in the local body segment coordinate system will be rotated into the global coordinate system using the measured changing orientation of the body segment (RBLG, Equation 4.11).
The body segments in the global coordinate system will then be connected together by their joint centres.
In Equation 4.11 RBLG describes the rotation from the body local to global coordinate system
at each epoch. RBLG is found from the instantaneous IMU orientation (RILG) and the
constant calibration matrix (RILBL) determined in equation Equation 4.10. Equation 4.11
4.3.2.
Experimental
Biomechanical analysis depends on human interpretation of data. Technology developments improve the amount of data available and its quality but the final analysis still rests with the biomechanist‟s interpretation of the results. To interpret the results of FMC it is necessary to understand the different sources of error. The difficulty of constructing an accurate body model is one source of error. The approximate size of this error and its effects on future biomechanical analyses needs to be determined.
The 3D anthropometric frame measurements of six subjects were compared against criterion anthropometric measurements taken by a certified technician. Version one of the frame was used (Figure 4.10). Seventeen different measurements were compared. In the anthropometric frame the 3D lengths were calculated as the length between two separately measured bony landmarks in 3D space. A qualified anthropometrist used a standard anthropometric kit to measure the lengths directly.
Table 4.2: Comparison of measurements between criterion anthropometry and our 3D anthropometric frame.
Comparison of Error Subject Error Mean Std Mean Relative CV
[cm] Error Dev Length Error
# Anatomical Reference 1 2 3 4 5 6 [cm] [cm] [cm] [%] [%] Right Side 1 Acromiale-radiale 1.5 1.3 1.1 0.9 0.8 1.2 1.1 0.3 33.6 3 1 2 Radiale-stylion 0.8 -1.8 -0.8 -1.2 0.8 -1.0 -0.6 1.1 25.7 -2 4 3 Iliospinale ht 8.3 7.0 8.7 7.5 2.0 4.3 6.3 2.6 98.4 6 3 4 Trochanterion ht -2.0 0.7 1.5 0.1 2.1 2.3 0.8 1.6 89.9 1 2 5 Trochanterion-tibiale laterale -12.9 -1.0 -0.6 -1.7 0.7 -0.1 -2.6 5.1 43.3 -6 12 6 Tibiale laterale ht 10.5 1.2 0.6 0.1 -0.1 0.8 2.2 4.1 48.0 5 9
7 Foot length (ak-pt) 0.6 -0.3 -0.4 -0.5 0.1 2.4 0.3 1.1 26.5 1 4
Left Side 8 Acromiale-radiale 4.0 0.7 0.6 -0.2 0.2 1.4 1.1 1.5 33.2 3 5 9 Radiale-stylion 2.5 0.6 -0.4 -1.6 -1.0 -1.1 -0.2 1.5 25.8 -1 6 10 Iliospinale ht 8.3 6.3 9.5 7.0 7.1 3.9 7.0 1.9 97.1 7 2 11 Trochanterion ht 0.7 0.3 0.8 0.8 1.0 1.7 0.9 0.5 88.6 1 1 12 Trochanterion-tibiale laterale -10.1 -0.6 -0.4 -0.5 -0.2 0.0 -2.0 4.0 43.3 -5 9 13 Tibiale laterale ht 9.7 -0.5 -1.2 -0.6 -0.8 0.2 1.1 4.2 47.1 2 9
14 Foot length (ak-pt) 0.1 0.0 -0.2 -0.5 -0.2 0.3 -0.1 0.3 26.5 0 1
Breadths 15 Biacromial breadth 6.8 1.9 0.5 3.1 3.2 2.6 3.0 2.1 39.6 8 5 16 Bi-iliocristal breadth 1.4 3.6 6.8 8.0 2.5 6.1 4.7 2.6 29.6 16 9 Height 17 Vertex Height -5.5 -11.7 -0.7 -0.1 -0.3 0.5 -3.0 4.8 174.7 -2 3 Column Mean 1.5 0.5 1.5 1.2 1.1 1.5 1.2 2.3 57.1 2.3 4.9
4.3.3.
Results and discussion
A method was presented previously to construct an athlete specific body model from 3D anthropometric measurements. A custom anthropometry frame was constructed and the accuracy of the frame measurements was tested by comparison to criterion measurements made by a certified anthropometrist. The body model is required in order to calculate the subject movements from the IMU data.
Creating a body model of the athlete with the 3D anthropometric frame was successful, but was a time consuming process, and it took an hour per subject. The process only needs to be completed once for each athlete. Additional motion capture sessions with the same athlete should only require a 5 second calibration of the IMUs while the athlete is constrained to a pre-measured reference position by the anthropometry frame measurement arms.
The calculated limb lengths for the athlete specific body model came from papers by Dumas and Reed (Dumas, et al., 2007; Reed, et al., 1999). The limb lengths were calculated from the position the joint centres, which were calculated from the underlying measurements of the bony landmarks. The possibility for error calculating limb lengths from bony landmarks was only discussed for the sacrum body segment by Dumas and Reed, which is probably, contains the most error because of the distance between the bony landmarks and the lumbar of hip joint centres. Based on an estimated vertical error in hip joint centre location of 3.5mm (the mean error from Reed‟s paper) and the mean sacrum segment length of 94 mm (from Dumas‟ paper) the estimated body segment limb length of 4% was chosen and adopted for all body segments. It was assumed this source of error would not affect the measured orientation of the body segments because orientation is generally calculated from the direction of vectors and not lengths.
The scaled inertial parameters used to create the athlete specific body model came from papers by Dumas and Reed (Dumas, et al., 2007; Reed, et al., 1999). The location of the body segment centre-of-mass and the inertia tensor were based on the characteristic dimensions of each body segment, which were inherently more accurate than some other approaches based on the regression of height and weight only. The underlying data came from McConville and Young and were based on 31 healthy males with a mean age of 27.5 years and a mean weight of 80.5 kg. McConville‟s mean age was about 5 years older than the mean age of the athletes and skiers have larger thighs than the average population. The inertial properties of the biomechanical man are therefore unlikely to be perfect for the athletes, but were the best estimate practically available. McConville used a stereo-photogrammetric technique that was reported to be within 6% error for CoM location measurements from cadavers, within 6% error for the principal moments of inertia and less than 10% error for body segment masses.