Minutes from RT Teleconference Dec. 11, 2008
- Workshop venue
- VK: FIT may be able to hold the meeting
- Fixed latency proposal: No new updates. Formal proposal
prepared after the new year.
- Evaluation Data: Preparing the Test Set
- Summary of promised data:
- AMIDA data:
- 1 focus group meeting
- 1 recording set of AMIDA meetings (4 meetings in
total). This data will have the remote participant.
- NIST has ~15 hours of focus group meetings to select from
- Potential data
- USC data: See the current RT email thread on this
matter. The data will contain head mics, centrally located table
mics (16-channel mic array + 1 table mic.), 5 video cameras.
- TH: Adam was checking on the availability of data from IDIAP.
- Action Item: JF will contact Adam and Herve
- Question for discussion: How to select the 3 hours of test
data: What's more important central mics or speaker variability?
- JF: speaker variability is important
- TH: a variety of meeting rooms are also important
- Leveraging multimodal/mic array output
Array Quality Assessment methods:
- NIST is starting to experiment with signal measurement methods