Natural Sounding Sub-Word Units Concatenation in Malay Speech Synthesis

摘要

The goal of this work was to concatenate Malay subwords without introducing perceptible audible discontinuities.Based on a phonemes adjacency analysis, we build a list of non-audible distortion sub-word unit lookup. Selecting subword from this lookup will avoid spectral mismatched between the joined speech units and this can guarantee the segmental quality of the sub-words concatenation. In order to avoid prosodic mismatched, we only allow the sub-words to be concatenated when they have the same prosodic features. The expectation by considering the correct prosodic features and the non-audible discontinuities of sub-word units, a natural sounding of sub-words concatenation is possible to be obtained.