The general themes of the conference include speaker and language recognition and characterization. The specific topics include, but are not limited to, the following:
- Speaker characterization and adaptation
- Features for speaker and language recognition
- Multi-speaker training, detection and diarization
- Robustness in channels and environment
- Robust classification and fusion
- Speaker recognition corpora and evaluation
- Speaker recognition with speech recognition
- Forensics, multimodality, and multimedia speaker recognition
- Speaker and language confidence estimation
- Language, dialect, and accent recognition
- Speaker synthesis and transformation
- Human recognition of speaker and language
- Analysis and countermeasures against spoofing attacks
- Commercial applications
REGULAR PAPER SUBMISSIONS
All regular submissions (max 8 pages) will be reviewed by at least three members of the scientific review committee.The regular submissions must include scientific or methodological novelty; the paper has to review the relevant prior work and state clearly the novelty in the Introduction part.The accepted papers will appear in electronic proceedings.
INDUSTRY TRACK AND DEMOS
Odyssey committee recognizes a large gap between theoretical research results and real-world deployment of the methods. To foster closer collaboration across industry and academia, Odyssey 2014 features an industry submission track. This can include a description of your target application, a product, a demonstrator, or any combination. In addition to voice biometrics providers, we encourage submissions from companies who are in need for speaker or language recognition technology.The industry paper submissions do NOT have to present methodological novelty, but the submission MUST address one or all of the following aspects:
- Description of the application, role of speaker/language recognition
- Research results and methods that worked well in your application
- Negative research results that have NOT worked in practice
- Unsolved problems 'out-in-the-wild' that deserve attention
NIST SPECIAL SESSIONS: I-VECTOR CHALLENGE & NIST SRE-2012 FOLLOW-UP
In addition to regular and industry paper submissions, Odyssey 2014 features two special sessions co-organized with National Institute of Standards and Technologies (NIST). NIST SRE-2012 special session focuses on extended analyses on the latest, NIST 2012 speaker recognition evaluation (SRE) benchmark, and is targeted for the participants of NIST SRE 2012.
The i-vector challenge is a new type of challenge targeted for anyone interested for a 'quick start-up' in speaker recognition. Building modern speaker and language recognition systems requires a lot of preprocessing, corpus engineering and computations, making it challenging for newcomers to enter the field. This prohibits piloting of possibly promising modeling ideas developed outside of speaker recognition community (e.g. machine learning and image processing communities). To bridge this gap, NIST organizes a new type of benchmark, i-vector challenge, synchronized with Odyssey 2014.
Preliminary due-date for paper submissions to both special sessions is February 2014. Submissions by this deadline will undergo review both by NIST and by the scientific review committee, and will be included to the conference proceedings if accepted. Late challenge submissions (without a paper) are also encouraged; they can be presented as posters, but will not undergo peer review nor will be included to the conference proceedings.
To ensure smooth organization, early (non-binding) preliminary sign-up is required. More details and registration for the i-vector challenge will be available in November 2013.
Odyssey 2014 features three awards:
- A best paper award
- A best student paper award
- Free registration for 1 to 2 top-performers in the i-vector challenge
Best 1 to 2 teams (max 1 person per site) in i-vector challenge are provided FREE REGISTRATION to the full Odyssey 2014 workshop.