Investigation of acoustic units for LVCSR systems

One important issue in designing state-of-the-art LVCSR systems is the choice of acoustic units. Context dependent (CD) phones re-main the dominant form of acoustic units. They can capture the co-articulatory effect in speech via explicit modelling. However, for other more complicated phonological processes, they rely on the im-plicitmodelling ability of the underlying statistical models. Alterna-tively, it is possible to construct acoustic models based on higher level linguistic units, for example, syllables, to explicitly capture these complex patterns. When sufficient training data is available, this approach may show an advantage over implicit acoustic mod-elling. In this paper a wide range of acoustic units are investigated to improve LVCSR system performance. Significant error rate gains up to 7.1 % relative (0.8 % abs.) were obtained on a state-of-the-art Mandarin Chinese broadcast audio recognition task using word and syllable position dependent triphone and quinphone models. 1