Suppose that we observed several time series and their underlying processes are discreted-valued. The sticky hidden-Markov model can be used to denoised these noisy time series. Run python3 shdp_demo.py to show a demonstration (with animation, see demo on Youtube).
Fox, E. B., Sudderth, E. B., Jordan, M. I., & Willsky, A. S. (2011). A sticky HDP-HMM with application to speaker diarization. The Annals of Applied Statistics, 1020-1056.