Signal automata and hidden Markov models

TL;DR

This paper introduces a generic method for efficiently inferring and updating hidden Markov models from time series data, enabling real-time model adaptation with constant-time updates.

Contribution

It presents a novel approach for inferring hidden Markov models that can be updated in constant time as new data arrives.

Findings

Efficient real-time model inference.

Constant-time updates for new measurements.

Applicable to various time series data.

Abstract

A generic method for inferring a dynamical hidden Markov model from a time series is proposed. Under reasonable hypothesis, the model is updated in constant time whenever a new measurement arrives.