I'm trying to implement map matching using Hidden Markov Models in Python.
The paper I'm basing my initial approach off of defines equations that generate their transition and emission probabilities for each state. These probabilities are unique to both the state and the measurement.
I'm trying to use Python HMM frameworks such as hmmlearn to emulate their findings, but all the libraries I've looked at only let you define an initial emission and transition matrix with the option to train it (and I don't think I need to train it if I have equations defining the matrices).
I'm looking at using the GaussianHMM in hmmlearn because my emissions are Gaussian, but I can't define an initial covariance and mean matrix because each emission has its own distribution (see equation 1 from the paper).
Additionally, each of the transition probabilities are dependent on the emissions (see eqn 2), so they can't be constant either.
Thanks for the help!
