The EMU Speech Database SystemEMU 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. Features in a nutshell
Emu/R Release 4.3, March 2012Emu/R 4.3 released. See Emu Release 2.3, November 2009Emu 2.3 is released. See | |||||||||
DownloadsBinary packages of Emu 2.3 are available for:
Note: this release works best with . All Emu downloads are available via the . EMU 2.4 Release, Upcoming soon 2012RELEASE NOTES Version 2.4 Bug fixes and asked features added. See http://sourceforge.net/projects/emu/support closed Bugs and Feature Requests. Some selected modifications are listed below:
EMU 2.3 Release, November 2009RELEASE NOTES Version 2.3 Bug fixes and asked features added. See http://sourceforge.net/projects/emu/support closed Bugs and Feature Requests. Some selected modifications are listed below:
Installation and Usage
TroubleshootingThis software is likely to contain bugs. If you think you have encountered a bug, please don't hesitate to consult our Bug Tracking System and eventually report your problem. PatchesDownload patches from: Emu's Sourceforge patch collection A patch fixes problems of the current version of the software. All patches will be included in the next software release. Please make sure that you install patches in the right chronological order. It might be, that some files are part of more than one patch. In this case, the lastest version includes earlier fixes already. Source DistributionThe source distribution is a snapshot of the Emu sources at the release date. The source should build easily on Linux with and should build on Solaris as long as a recent gcc and Gnu make are available. It may also build on other flavours of Unix but has only been tested on these two platforms. Instructions for compilation are in the README file within the distribution. In addition to the above released sources, the latest version of the source via anonymous CVS is available from the . Please note that the CVS version may not compile in all configurations. A number of packages needed to build a full EMU system are version Version 8.5 of and the latest versions (CVS snapshots) of libassp and tclassp, both found at libassp Emu also requires the Tcl standard library (tcllib), tklib the Bwidgets Tcl library, both available via Sourceforge as well as the memchan, tclvfs, trf and tkimg packages. All these packages are also available via Active Tcl. | |||||||||
Library Emu in the Software package RThe Emu-R library is a collection of programs running within the R programming language for the analysis of any speech database that can be read by the Emu Speech Database System. Emu/R 4.3 released - soon available on CRAN (R)
Download .
R> install.packages("/Path/to/emu_4.3.tar.gz", repos=NULL, type="source") This release is an update due to new R 2.14 features and requirements. Emu/R 4.2 releasedThis release works best with . RELEASE NOTES Version 4.2
Emu/R 4.1 released This release works best with .
Installing Emu/RInstallation of the Emu functions for R is simple as the package is now hosted by the The Comprehensive R Archive Network.
Alternatively, you can dowload the Emu-R package 4.2 from SourceForge:
The actual R-CRAN package (4.2) requires >= EMU2.3 | |||||||||
Emu and ToBIToBI is a system for transcribing the intonation patterns of spoken language. ToBI defines a number of annotation levels and the criteria for placing labels on each. These labels include a segmentation into words along with tone labels which mark prosodic events such as prosodic tones and phrase boundary events. ToBI annotations have largely been made using the Unix based Waves+ toolkit from Entropic and the example materials are made available in the ESPS format which is readable only by Waves+ (and by Emu when an ESPS licence is available -- ie. on a Unix platform). This page describes some tools for using ToBI annotation in the Emu system. The ToBI training materials available from the Ohio State web site are in the ESPS format. We have converted these files to the SSFF format read by Emu on both Unix and Windows. These files are now available on our server:
The smaller files consist of only 12 utterances from the database (those beginning with `a'). These packages contain the original label files (augmented with a dummy label at the start of the word level so that Emu can treat the words as segments rather than events). Two Emu template files are provided, one mimics the traditional ToBI annotation scheme which presents four independant tiers, the other adds domination relations and two additional levels for intonational and intermediate phrases. A script is provided to convert traditional flat ToBI annotations into hierarchical ones. An example hierarchical annotation is shown below.
In this scheme, the Tone level is preserved, non-phrasal tones are linked to the word in which they occur and words are grouped into Intonational and Intermediate phrases based on the position of phrase boundary events. Pitch Tracking on WindowsResearchers investigating prosody have relied on ESPS/Waves+ for both labelling and pitch tracking. Emu can manage the labelling role but does not provide a pitch tracker. There are a number of possible pitch trackers that might be of use. Fortunately, the ESPS codebase has now been donated to the KTH speech group and is now being integrated into the Snack toolkit. The most recent release of Snack includes the ESPS pitch and formant tracker code and the most recent Emu release contains a simple tool to run these over speech data and produce SSFF formatted data files for your corpus. | |||||||||
Publications
| |||||||||
Emu and MySQL | |||||||||
Frequently Asked QuestionsThis is a very short list of FAQ. If you have any problem, see the Emu forums especially the Help forum. Your problem might be a software bug. In this case see Emus Bug tracking system. How do I get started with the EMU speech database system?
How can I get access to my existing databases?
Are there some patches?Download patches from: Emu's Sourceforge patch collection A patch fixes problems of the current version of the software. All patches will be included in the next software release. Please make sure that you install patches in the right chronological order. It might be, that some files are part of more than one patch. In this case, the lastest version includes earlier fixes already. Convert Labels freezes converting Praat TextGrid filesThis problem occurs when the TextGrid file is UTF-16 encoded. Until EMU support UTF-16, change the encoding to UTF-8. Therefore change Praat preferences (open and save TextGrid again) or use a text editor to convert.
| |||||||||
Miscellaneous Information
We are in the process of transferring additional information from the old
website to the new design. Meanwhile please visit the old site:
Steve Cassidy and the Emu of EMU | |||||||||
DocumentationsDocumentations of the Speech Tools of EMU Version 2 are included in the tools by a file menu entry Help or by the context menu below the emu picture. For other help see Documentation of EMU 1.7 Video documentationWatch the according to Harrington(2010). Harrington, J. (2010). The Phonetic Analysis of Speech Corpora. Blackwell. See older video tutorials below. | |||||||||
How do I install the EMU speech database system?
|
|||||||||
Emu and MySQLI've recently carried out some experiments involving converting an Emu hierarchical annotation into tables for a relational database. The point of this exercise is twofold. Firstly to look for improvements in the speed of searching large databases for segments. Secondly to demonstrate the equivalence of the Emu hierarchical model to the relational model, which is much better understood. The results of the experiments are summarised in a paper to be presented at Eurospeech99 in Budapest. Database Software: MySQLInitial experiments were carried out with the PostgreSQL system which is included with RedHat Linux 5.2. This system proved easy to use but gave very slow response times to queries: around five minutes for the example domination query. Some research turned up some comments about PostgreSQL not being a very fast system and so MySQL was tried as an alternative. Initial results were very encouraging and so far MySQL has shown to be as fast or faster than Emu on all queries tested. MySQL is available for most Unix systems and for Windows systems (although a fee is payable for Windows users). Scripts are given here which convert an Emu database to ASCII tables and a set of instructions to import the tables into MySQL. Although I have no experience with other database systems I imagine that the scripts could be adopted to import the data into any relational database system (eg. Oracle, Microsoft SQL Server). ScriptsThe conversion script emu2dbase (a downloadable version is here) is written in Tcl using the Emu extensions. The script takes two parameters, a template name and a pattern matching utterances to be translated into the relational tables. A directory is created named after the template and four files are created in the directory:
The script can be given an flag emu2dbase children 'sp1:*' this will convert all utterances matching the pattern
The database, with the same name as the template file, is created by running MySQL with the database file as input: mysql < database this will delete any existing database, create a new one and create the appropriate tables for tokens, links and levels. These tables are then populated from the text files. The database can then be queried as per the examples in the paper. ResultsThe results with MySQL are impressive, simple queries are very fast and even sequence and domination queries run in less time than the standard Emu query engine. Since the Emu engine searches all utterances sequentially, it will not scale well to very large databases. The SQL engine uses indexes to improve efficiency and should scale well, although further experimentation is required to verify this. Further developments will be posted here as we are able to do more experiments. Any comments are welcome. |
Copyright ©
2010, Institute of Phonetics and Speech Processing, LMU Munich
Copyright ©
2006, Institute of Phonetics and Digital Speech Processing, CAU Kiel
Copyright ©
2001, Department of Linguistics, Macquarie
University.
Last update 25 May 2010