I want to train a Hidden Markov Model using the python library hmmlearn . For this, I have a dataset which consists of multiple experimental runs (different people perform series of actions like "sitting", "staying" and "lying") and the observations of each run are generated by sensors attached to the person. My question now is: how to train the model considering the fact that my observation sequences have been derived from different runs? Should I introduce additional labels like "experiment_start" and "experiment_stop" in order to be able to merge several runs in my train set? Thank you!