Asynchronous-transition HMM

摘要: We propose a new class of hidden Markov model (HMM) called asynchronous-transition HMM (AT-HMM). Opposed to conventional HMMs where state transition occurs simultaneously all features, the allows transitions asynchronized between individual features better asynchronous timings acoustic feature changes. In this paper, we focus on particular AT-HMM with sequential constraints based novel concept "state tying along time". To maximize advantage model, also introduce feature-wise technique. Speaker-dependent speech recognition experiments demonstrated error reduction rates more than 30% and 50% in phoneme isolated word recognitions, respectively, compared HMMs.