Hidden Markov Model Explainer
A hidden Markov model, or HMM, represents a sequence whose underlying state cannot be observed directly. Each hidden state emits observable evidence, and transition probabilities describe how the state is likely to change from one step to the next.
The Markov assumption makes the next state depend on the current state rather than the entire earlier history. The observation model describes how likely each visible signal is under each possible hidden state.
Algorithms can calculate the likelihood of an observed sequence, infer the most probable hidden-state path, or estimate parameters from examples. HMMs have supported speech recognition, biological sequence analysis, activity recognition, and equipment monitoring.
The assumptions can be too restrictive when durations or long-range dependencies matter. State meanings may also be ambiguous, so model selection, domain validation, and comparisons with richer sequential methods are necessary.
A hidden Markov model describes a sequence generated by unobserved states that change over time and produce observable signals.
Key terms include hidden state, observation, transition probability, emission probability, Markov assumption, forward algorithm, Viterbi algorithm and sequence.
Transitions describe how an unseen process changes, while emissions connect each hidden state to observable evidence.
Inference works backward from the visible sequence to probabilities over the states that may have produced it.
The standard model assumes the current state summarizes the relevant past for predicting the next state.
Longer dependencies or explicit state durations may require extensions or a different sequence model.
Explore more "Explainers"
Discover additional explainers across politics, science, business, technology, and other fields. Each explainer breaks down a complex idea into clear, everyday language—helping you better understand how major concepts, systems, and debates shape the world around us.
