Skip to Content

Department of Linguistics

The MU-Talk Text-to-Speech System

Robert Mannell

The MU-Talk Text-to-Speech system developed in the Speech, Hearing and Language Research Centre (SHLRC) at Macquarie University, has evolved gradually since the work by John Clark in the mid 1970's on the SID serial formant synthesiser and its accompanying synthesis-by-rule based text-to-speech system. In the mid 1980's, work by Clark, Summerfield and Mannell substituted the SID synthesiser with a parallel formant synthesiser. This was followed by the development by Clark and Mannell of an accompanying SBR-based TTS system targeting this parallel formant synthesiser. Up to this time, the systems were all written in Fortran, but all subsequent systems have been written in C and, more recently, in C++. In the early 1990's Clark and Mannell began integrating a diphone concatenation system into the SHLRC TTS system. There were two strands of this work, one examining parallel formant parameter diphone concatenation and the other examining channel vocoder parameter diphone concatenation. Since the mid-1990's Mannell has worked sporadically on this system. This work has involved the complete restructuring of the system. The system, in its current configuration is now known as MU-Talk (where "MU" stands for Macquarie University).

On the pages linked to this document, the current status of MU-Talk is outlined and current work and future plans are also examined.

Figure 1: The MU-Talk Text-to-Speech System