Published June 19, 2017 | Version v1
Dataset Open

ZS2017 AaltoLAG Submission #1 (mostly testing)

  • 1. Aalto University

Description

First submission to the IEEE ASRU / Zero-Resource Speech Processing Challenge with a syllable-based pattern discovery system. Mostly testing the functionality of the system and ensuring that track1 / track2 results replicate for Mandarin/English/French also on the server. 

The track1/track2 systems require MATLAB (see the readme in /system/ for dependencies and tested platform).

Questions and comments to okko.rasanen@aalto.fi.

Note: This package is NOT an official distribution of the included pattern discovery system. The package is simply intended as a ZS2017 challenge submission and the current version contains large amounts of undocumented lab code. 

Notes

See the readme in the system folder for a brief description of the system. Assumes cosine distance and pattern length normalization in track1 evaluation. Please see the readme in the system folder for a list of dependencies that are included in the package for completeness of challenge submission, but are not created or maintained by the present authors.

Files

Files (3.7 GB)

Name Size Download all
md5:4d1bf3e93d5d32e926b6e80bc1d7a5c5
3.7 GB Download