Many problems of academic and industrial interests such as operating mode switching, feed quality change, process fault diagnosis and phase change analysis in a chemical process either take a noncontinuum form or can be represented by a discrete-valued state and be readily solved. In this study, a noncontinuum state estimator, which is based on a moving horizon method for a class of switching system that follows a hidden Markov model, is developed. A new approach to arrival cost development and constraints handling in moving horizon estimation of noncontinuum state are discussed. The effectiveness of the proposed method is illustrated by considering mode estimation problems in a simulated leakage detection system as well as a continuous stirred tank reactor.