Institute of Phonetics and Speech Processing
print

Links and Functions
Language Selection

Breadcrumb Navigation


Content

MAUS - Munich AUtomatic Segmentation

The aim of the project Munich AUtomatic Segmentation (MAUS) is the development of fully automatic algorithms that segment and label a speech signal into its phonetic segments.

MAUS uses a combination of machine-learned and statistically weighted rules to predict possible pronunciation variants, and a HTK based HMM search algorithm to find the most likely segmentation and labelling.
MAUS is free-ware and can be down-loaded from our FTP server (see links) or used as a web-service (under construction within the CLARIN-D project). To use MAUS on your computer you'll need the following requirements:

  • Linux based OS, including csh and gawk
  • HTK
  • sox

Running MAUS on Windows is possible, but the installation is tedious since the MAUS package consists of csh scripts (you'll need cygwin to run these under Windows) and c++ code (you need a compiler for c++).

MAUS currently supports the languages German (full), English, Hungarian, Islandic, Estonian, Italian (without pronunciation modelling), Spanish, Portuguese (beta). MAUS can be adapted to new languages on several levels of performance (contact F. Schiel if you plan to do so).

 Links

Original MAUS Website

Download MAUS

Florian Schiel

Publications

Conference papers

Dissertations

  • Kipp, A. : Automatische Segmentierung und Etikettierung von Spontansprache; Shaker Verlag Aachen 1999.

  • Beringer, N. : Regeladaptive kategoriale Analyse von Spontansprache; Shaker Aachen 2002