Viterbi Algorithm

Generalising the above expression, the probability of the partial best path to a state i at time t when the observation kt is seen, is :

Here, we are assuming knowledge of the previous state, using the transition probabilites and multiplying by the appropriate observation probability. We then select the maximum such.