Design of a speech recognition system based on non-uniform segmental units

The design of speech recognition system based on acous-tically-derived, segmental units can be divided in three steps: unit design, lexicon building and pronunciation mod-eling. We formulate an iterative unit design procedure which consistently uses a maximum likelihood (ML) objec-tive in successive application of resegmentation and model re-estimation. The lexicon building allows multi-word en-tries in the lexicon but restricts the number of these entries in order to avoid a too costly search. Selected multi-word lexical entries are those with high frequency (such as func-tion words) and those which consistently exhibit cross-word phone assimilation. The stochastic pronunciation model represents the likelihood of a particular acoustic segment se-quence given the phonetic baseform of a lexical item, where the sequence of baseform phones are treated as a Markov state sequence and each state can emit multiple segments. 1