How to make gate predictions for certain movements using HMM model?

Question

I want to use IMU sensors to predict which gate the user is going to move into next (for walking, running, jumping, etc). From what I've researched it seems like HMM is the best machine learning model for state predictions. But all the models I am seeing online are based of pre-recorded static data instead of live data that is being recorded. How should I go about implementing this?

I have already implemented a rolling window that uses the last 30 data records captured to make the next prediction and have implemented normalisation on them. I have identified the gates I want to recognise for each movement and I have a classification algorithm already set up that goes through each window and uses it to classify the current movement. I was planning to use these to validate the predictions I will be making (or even to use it as input to predict the next movement of the person if needed).