Recognition of facial expressions - Knowledge Based Systems ...

Equation 18
z
x&
x&
x&
1
2
3
0 1 0
= 0 0 1
0 0 − β
2
= [ 2σ
β 0 0]
x
x
x
x
x
x
1
2
3
1
2
3
+
0
0
1
u(
t)
In the model we use, the state vector contains an additional state variable according to the
Gauss-Markov process. u(t) is a unity Gaussian white noise. The discrete form of the
f t
model for tracking the eyes in the sequence is given in Equation 17. φ = e ∆ , w are the
process Gaussian white noise and ν is the measurement Gaussian white noise.
Equation 19
x
z
k
k
= φ + w
k
= H
k
⋅ x
k
k
+ v
k
The Kalman filter method used for tracking the eyes presents a high efficiency by
reducing the error of the coordinate estimation task. In addition to that, the process does
not require a high processor load and a real time implementation was possible.
- 57 -