Hidden Markov Model
A probabilistic sequence model with unobserved states that evolve according to Markov transitions and emit observable values according to state-dependent distributions.
Machine-Learning Context
A Hidden Markov Model represents a sequence with latent states connected by Markov transitions and state-dependent observation distributions. Classical ASR systems commonly combined HMM state sequences with Gaussian or later neural acoustic emissions; forward-backward and Viterbi algorithms provide core inference procedures.
Modeling Boundary
An HMM summarizes relevant history through the current hidden state under the Markov assumption. That is a different dependency model from modern neural sequence architectures that can condition on much longer context directly.
Related Machine-Learning Concepts
- Viterbi Algorithm
- Automatic Speech Recognition
- Markov Chain
- Dynamic Programming