[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[AUDITORY] COVAREP - Call for contributions

CALL for contributions

We are pleased to announce the creation of an open-source repository of advanced speech processing algorithms called COVAREP (A Cooperative Voice Analysis Repository for Speech Technologies). COVAREP has been created as a GitHub project (https://github.com/covarep/covarep) where researchers in speech processing can store original implementations of published algorithms.

Over the past few decades a vast array of advanced speech processing algorithms have been developed, often offering significant improvements over the existing state-of-the-art. Such algorithms can have a reasonably high degree of complexity and, hence, can be difficult to accurately re-implement based on article descriptions. Another issue is the so-called 'bug magnet effect' with re-implementations frequently having significant differences from the original. The consequence of all this has been that many promising developments have been under-exploited or discarded, with researchers tending to stick to conventional analysis methods.

By developing the COVAREP repository we are hoping to address this by encouraging authors to include original implementations of their algorithms, thus resulting in a single de facto version for the speech community to refer to.

We envisage a range of benefits to the repository:
1) Reproducible research: COVAREP will allow fairer comparison of algorithms in published articles.
2) Encouraged usage: the free availability of these algorithms will encourage researchers from a wide range of speech-related disciplines (both in academia and industry) to exploit them for their own applications.
3) Feedback: as a GitHub project users will be able to offer comments on algorithms, report bugs, suggest improvements etc.

We welcome contributions from a wide range of speech processing areas, including (but not limited to): Speech analysis, synthesis, conversion, transformation, enhancement, speech quality, glottal source/voice quality analysis, etc.

In order to achieve a reasonable standard of consistency and homogeneity across algorithms we have compiled a list of requirements for prospective contributors to the repository. However, we intend the list of the requirements not to be so strict as to discourage contributions.

Getting contributing institutions to agree to a homogenous IP policy would be close to impossible. As a result COVAREP is a repository and not a toolbox, and each algorithm will have its own licence associated with it. Though flexible to different licence types, contributions will need to have a licence which is compatible with the repository, i.e. {GPL, LGPL, X11, Apache, MIT} or similar. We would encourage contributors to try to obtain LGPL licences from their institutions in order to be more industry friendly.

We believe that the COVAREP repository has a great potential benefit to the speech research community and we hope that you will consider contributing your published algorithms to it. If you have any questions, comments issues etc regarding COVAREP please contact us on one of the email addresses below. Please forward this email to others who may be interested.

Existing contributions include: algorithms for spectral envelope modelling, adaptive sinusoidal modelling, fundamental frequncy/voicing decision/glottal closure instant detection algorithms, methods for detecting non-modal phonation types etc.

Gilles Degottex <degottex@xxxxxxxxxx>, John Kane <kanejo@xxxxxx>, Thomas Drugman <thomas.drugman@xxxxxxxxxxx>, Tuomo Raitio <tuomo.raitio@xxxxxxxx>, Stefan Scherer <scherer@xxxxxxxxxxx>

Website - http://covarep.github.io/covarep 
GitHub  - https://github.com/covarep/covarep