The EMU Speech Database System

EMU is a collection of software tools for the creation, manipulation and analysis of speech databases. At the core of EMU is a database search engine which allows the researcher to find various speech segments based on the sequential and hierarchical structure of the utterances in which they occur. EMU includes an interactive labeller which can display spectrograms and other speech waveforms, and which allows the creation of hierarchical, as well as sequential, labels for a speech utterance.

Emu 1.10.5 Released

This version is primarily a bug fix (bug item #1338464) release.

Emu 1.10 Released

This version, again, adds a number of new tools and improves some existing ones. The major enhancements are:

  • New graphical query GUI
  • y-scalability of waveforms
  • SignalViewLevels provides a way to hide or display levels in signal view
  • Segment / event list handling in tkassp improved: Source track can now be chosen

Emu 1.10 also fixes some minor bugs. Upgrade from previous versions should be smooth.

Emu 1.9 Released

The version adds a number of new tools and improves some existing ones. Emu now includes two tools for deriving pitch, formant and other tracks from speech signals. The suite of graphical tools available in Emu now includes:

  • AutoBuild tool: supports running autobuild scripts on many utterances in a database.
  • Template Editor: a graphical template editor which helps create and modify template files ensuring that they are properly formed.
  • Speech Signal Analysis (Tkassp): derive pitch, formant, rms and other tracks from source audio files using Michel Scheffer's Assp tools. Save results in SSFF format for use in Emu.
  • Pitch & Formant Tool: derive pitch and formant tracks using the Snack (formerly ESPS) algorithms. Save results in SSFF format for use in Emu.
  • Recording Tool: a tool to assist in recording from a list of prompts. Each recording is saved in a seperate file and a template file can be constructed to work with the data.
  • Emu Segmenter: a tool to help segment large audio files into smaller chunks for analysis. Able to load word lists and automatically segment the signal based on periods of silence.

The release also includes a large number of bug fixes and minor improvements in existing tools and should be a stable upgrade from the earlier version.

Emu Homepage Moved

The 'official' home page of Emu is now this Sourceforge page (http://emu.sourceforge.net/). The move reflects the fact that the project is now an International collaboration rather than being based within SHLRC at Macquarie University. The old website will be forwarded but please update any links that you may have to our site.

Emu 1.8 Released

This is a bugfix release however one significant addition is that of a formant and pitch tracking tool based on the recently release Snack tools. These are based on the same code as the old ESPS formant and pitch trackers and so should be useful tools. The advantage of the Snack versions is that they run on Windows and Macintosh as well as Unix and no license is needed to run them.

You can download the release here and read the release notes here.

Emu 1.7 includes Macintosh Support

The new 1.7 release of Emu includes support for the Apple Macintosh for the first time. The downloads page has details of how to download and install Emu on the Mac but here is a small screenshot just to prove that it works:

The screenshot also shows the new Emu query tool which was built to allow querying and data extraction on the Mac but which should also prove useful on other platforms. Since R is also available for the Macintosh, users of this platform can now enjoy the full Emu experience!

Planning for Emu 2.x

We are putting together a document outlining our plans for Emu development beyond the current system. The document is available here. Please feel free to address any comments to Steve Cassidy

Emu Development moves to Sourceforge

I have recently begun moving Emu development to SourceForge which is a site supporting Open Source software development. SourceForge provides a number of useful facilities such as a bug tracking system and a way of logging and tracking support requests. The Emu mailing list will also be moved to the new site. This page will still be the definitive Emu homepage but users interested following the development of Emu should find the new site useful. Please see The Emu page at SourceForge for more information.

Emu and Splus/R

These are extensions to the Splus and R statistical packages which provide an interface to Emu functions and many special purpose functions for analysing and visualising speech data. While this package predates Emu (it is the major part of the earlier MU+ system), they have not yet been made widely available. I have now managed to package them up and provided installation instructions.

R is a freely available system, Splus is commercial software. Both systems run on Unix and Windows platforms. The Emu library is compatable with all versions and platforms. For more information see the Emu/Splus page.

Emu and ToBI

Emu has been in use for some time to mark up ToBI style prosodic annotations. With the help of Mary Beckman at Ohio State we have put together a version of the English ToBI training materials in SSFF format which can be read by Emu on Windows and Unix platforms. See the Emu ToBI page for details of how to download this dataset.

Download Documentation

You can now download the entire Emu manual as a zip file emu-doc.zip.

Emu and Relational Databases

I've recently completed a paper for Eurospeech99 which looks at compiling Emu annotations into the relational model. The paper is available online as are some comments about the experiments which include the scripts used to generate the relational table.

Publications

We have recently had a paper accepted for a special issue of Speech Communication on Linguistic Annotation. Other papers have been presented at Eurospeech 99 and the Australian Database Conference, 2000. See my publications page for detailed references.

Emu is included on a CDROM accompanying our book, Techniques in Speech Acoustics which was published in May 1999 by Kluwer.

Linguistic Annotation

Steven Bird and Mark Liberman have been collating a list of linguistic annotation methods and tools. One of their goals is to bring together the many different annotation tools and formats and define a useful set of standards that such tools might work with. Emu goes some way towards being able to deal with different kinds of annotation; as an example of using Emu with other annotation formats I've written some thoughts about Emu and the BAS Partitur label format, comments are welcome.

Mailing List

Two mailing lists (emu-announce and emu-devel) are maintained on the Emu developers site at SourceForge. Please see that site for details on how to subscribe.


For more information, please send mail to Steve.Cassidy@mq.edu.au.

Copyright © 2001, Department of Linguistics, Macquarie University.