Peter Cahill

Exemplar-based pitch contour generation using DOP for syntactic tree decomposition

The generation of a pitch contour from linguistic information has long been recognised as a requirement for natural sounding speech synthesis. This paper investigates the use of an exemplar-based model for pitch contour generation. The main drawbacks of previous unit selection-based approaches for pitch contour generation is determining the size of the unit, and to guarantee that only prosodic and linguistically related units will be selected. The work presented in this paper overcomes these drawbacks by using only prosodic-syntactic correlated data, and a dynamic unit size model using data-oriented parsing. An AB comparison perceptual test showed 58% preference for the exemplar-based model, 25% for a HTS model, and 17% find both the same in terms of naturalness and pitch. In a MOS test, exemplar-based model achieved higher scores than that the HTS model achieved.

Muse: An open source speech technology research platform

This paper introduces the open source muster speech engine (Muse) for speech technology research. The Muse platform abstracts common data types and software as used by speech technology researchers. It is designed to assist researchers in making repeatable experiments that are not hard coded to a specific platform, language, algorithm, or corpus. It contains a script language and a shell where users can interact with various components. The presentation of this paper will be accompanied by a demo at the SLT workshop.

Pitch recovery of missing syllables using sparse representation in exemplar-based pitch generation

The use of exemplar-based techniques for pitch generation in a text-to-speech system has shown a high degree of success and very comparable results compared to other techniques. The use of these techniques, however, requires that all units occur in the corpus. One of the limitations of this requirement is that the prosodically correlated data to the input found in the corpus does not always contain suitable units, and sometimes no units could be found in the corpus. These non-existent units can be seen as missing parts from the pitch signal. The work presented in this paper overcomes the missing units problem by using sparse representations for missing pitch data recovery. The framework proposed works in two stages; the first stage uses a unit selection approach to generate the initial pitch contour, the second stage adopts a sparse representation to generate the pitch contour for the missing units identified in the first stage. The approach followed showed comparable results compared to other pitch generation methods.