Biomedical Engineering Reference
In-Depth Information
TaBlE 6.1:
Comparison results of all algorithms with different
Q
k
MSE dIag(
Q
K
)
adaPTIVE FIlTERINg oF
PoINT PRoCESS
SEQUENTIal ESTIMaTIoN
MLE
CoLLapsE
[2e-5, 1e-7]
0.04801
0.1199
0.04522
[2e-5, 1e-6]
0.1081
0.1082
0.0489
[1e-5, 1e-6]
0.1076
0.1013
0.0588
assumption (solid gray line).The performance difference can be attributed to a more accurate esti-
mate of the real posterior density because no further assumptions are made. The MLE estimation
does not perform well in this simulation for reasons that are not totally clear, but may be related
to the multiple maxima of the posterior. As an example in Figure
6.4
, the posterior density at time
45.092 sec is shown (blue solid line) with three ripples and is obviously not Gaussian distributed.
velocity reconstruction
desired velocity
velocity by seq. estimation (MLE)
velocity by seq. estimation (collapse)
velocity by adaptive filtering
1.5
1
0.5
0
-0.5
-1
-1.5
2
3
4
5
6
7
8
9
10
x 10
4
time (ms)
FIgURE 6.3:
Velocity reconstruction by different algorithms.
