A novel model for phoneme recognition using phonetically derived features

This paper presents work on the use of segmental modelling and phonetic features for phoneme based speech recognition. The motivation for the work is to lessen the effects of the IID assumption in HMM based recognition. The use of phonetic features which are derived across the duration of a phonetic segment is discussed. In conjunction with the use of these features, a hybrid phoneme model is introduced. In a classification task on the TIMIT database, these features are capable of outperforming standard HMM. The extension of the work to recognition is presented in detail. The challenges are identified and a novel algorithm presented for recognition based on phonetic features and the hybrid phoneme model. The approach is built around a segmentation hypothesis approach employing pruning at a number of levels.