Classification

• Use Bayes’ rule:

P(Ui | F0..t H0..t C0..t) = P(F0..t H0..t C0..t | Ui) P(Ui) / P(F0..t H0..t C0..t )

• Assume independence of modality and time

  P(F0..t H0..t C0..t | Ui ) = P(F0..t | Ui ) P(H0..t | Ui ) P(C0..t | Ui )
  
  P(F0..t| Ui ) = P i P(Fj | Ui )

• Individual observations P(Xj | Ui ) modeled with empirically derived approximate Gaussian densities.

Previous slide

Next slide

Back to first slide

View graphic version