# What is meant by decoding in a Hidden Markov Model?

HMM contains two types of states: observable and hidden. Let $$\{ h_1,h_2,h_3,\cdots,h_n\}$$ be hidden states and $$\{o_1,o_2,o_3,\cdots, o_m\}$$ be the observable states.

Suppose the $$n^2$$ transition probabilities $$p(h_j|h_i)$$ and the $$mn$$ emission probabilities $$p(o_j/h_j)$$ are given along with the initial probability distribution vector $$\pi =[\pi_1, \pi_2, \pi_3, \cdots, \pi_n]$$

Then what is meant by decoding in HMM?