The likelihood function is the joint density of the observations i.e.

L(θ) = f(Y) Now consider instead the joint density of both states X and observations Y. This can be written as

f(Y) = ∫_Xf(X, Y) dX = ∫_Xexp log f(X, Y) dX