Continuous wavelet transform as a tool for prosody research

The fact that prosody is hierarchical in nature is implicit in most prosodic theories and models, be they of phonological or phonetic origin. However, the theories and models have relied on signal processing and representations that lack — in most part — the affordances for studying the hierarchy directly. Therefore, different models have been developed, for instance, to represent intonation, breaks, and prominences. This has left prosodic studies without a common framework for talking about the relevant phenomena arising from the suprasegmental variables of speech signals. In my talk, I will present a new, continuous wavelet transform (CWT) based tool we have developed in our research group at the University of Helsinki. The talk will shortly describe the CWT as a method for deriving a scale-space representation of speech signals and its use in basic phonetic research as well as speech synthesis.