The Emu system is the current incarnation of a software system that began life at Edinburgh University in the late 1980s by Jonathan Harrington and Gordon Watson as APS (Acoustic Phonetics in S). This early system consisted of a set of extensions to the S statistical package running on Unix systems which allowed queries to be made of speech databases with the resulting data being read into S for analysis and visualisation.
While APS allowed only one dimensional labels to be placed on speech utterances, the subsequent development of the MU+ system by Jonathan Harrington and others at Macquarie University generalised the labelling structure to multiple interconnected labelling tiers. For the first time it was possible to query a speech database on both hierarchical (phonemes which are part of strong syllables) and sequential (vowel phonemes following stops) constraints. The MU+ system was still based on the Splus language but now provided a rich set of visualisation and analysis tools developed specifically for speech analysis. The system was developed in the context of the Australian National Database of Spoken Language but was used for numerous smaller projects within SHLRC at Macquarie University. MU+ included a tool (labed) for automatically generating hierarchical descriptions of utterances based on various sets of rules for syllabification, phonetic to phoneme mappings etc. Much of the programming work for the MU+ system was carried out by Jonathan Harrington and Andrew McVeigh. MU+ was eventually ported to the Windows 95 version of Splus and was used by a number of laboratories around the world for speech database analysis.
The Emu system is a re-implementation of the core of MU+ which generalises the concept of multi-tiered labelling and provides a set of portable, standalone tools as well as the existing Splus based interface. One motivation for the new implementation was the difficulties we were having applying MU+ to other languages, such as Mandarin. Another need was for our own graphical labelling system for use by students: earlier systems had used the commercial Waves+ system from Entropic which was too expensive to use in teaching. Starting from scratch enabled us to remove some of the arbitrary constraints inherent in the MU+ system and provide an easily configurable system which was flexible enough to adapt to any speech database project. Most of the programming work for Emu was done by Steve Cassidy but Emu retains a number of components from MU+ including most of the Splus code for visualisation and analysis of speech data. A new interface to the freely available XlispStat system was also built to parallel the facilities in Splus, this system is now used extensively in teaching speech science at undergraduate and graduate levels.
The contributors to Emu are: Steve Cassidy, Jonathan Harrington, Andrew McVeigh and Catherine Watson. Emu uses a number of software components from other groups, in particular Tcl/Tk from Scriptics, and the Edinburgh Speech Tools library from CSTR at Edinburgh University.