First, we will introduce the hidden Markov model, which is a probability model of time series. The concept book has its original meaning: an unobserved state random sequence randomly generated by a hidden Markov chain, each State generates an observed random sequence. Combined with the example in the book, this state can be a box. The observed values are the red or white balls in the box.
The following are some definitions:
Q: the set of all possible states q = {Q1, q2,..., qN} n is the number of possible states.
V: all possible observations v = {V1, V2,... VM} m are possible observations.
I: The state sequence O with a length of T is the corresponding observation sequence.
I = (I1, I2,... it), O = (O1, O2,..., ot)
A is the state transfer probability matrix:
A = [AIJ] nxn (1)
Where,
AIJ = P (IT + 1 = Qj | it = Qi), I = 1, 2,... n; j = 1, 2,..., n (2)
It indicates the probability that the Qj is transferred to the State at the t + 1 moment under the condition of T and State Qi.
B is the observed probability matrix:
B = [bij] nxm (3)
Where,
BJ (K) = P (Ot = VK | it = Qj), k = 1, 2,... m; j = 1, 2,..., n (4)
It indicates the probability of generating a observed VK under the condition of T and status Qj.
Is the initial test state probability vector:
(5)
Where,
(6)
It is the probability that T = 1 is in the State Qi.
Therefore, the hidden Markov model is determined by the preceding three elements:
(7)
The following example describes the three elements. (Used directly)
Experiences of hidden Markov models in statistical learning methods