FACE 2004 SMC

1
FACE 2004 SMC

• Zheng-Hua Tan and Paul Dalsgaard
• It is two years now since FACE project started in
  January 2002. What is NEXT?

2
In the past

• In the context of automatic speech recognition in
  mobile communications, SMC activities are shaped
  by
• channel error protection scheme
• packet loss concealment
• transmission error concealment
• exploiting the knowledge about transmission
  errors for the purpose of optimising the OOV
  detection.

3
In the past cont.

• Besides, we have a number of field trails in the
  area of speech enabled access to network hosted
  applications
• personal assistant
• vehicle registration database on-line query
  service
• medical application.

4
In the year of 2004

• The FACE activity 2004 - SMC will definitely
  continue in similar lines and will combine with
  other activities which are going on at SMC mainly
  CNTK and MAGNET.

5
In the year of 2004 cont.

• The general goals are in two aspects
• 1. robust speech recognition in the context of
  mobile communications
• To incorporate speech recognition into
  real-world application, robustness is a key issue
  to be solved. In mobile communications, channel
  noises (transmission errors and packet losses)
  are additional source for degradation of speech
  recognition accuracy besides environmental/acousti
  c noise. The combination of these two noises
  makes the problem even more severely.

6
In the year of 2004 cont.

•     2. speech enabled applications
• Speech technology has come to the edge between
  laboratory research and real-world application.
  We will continue our effort on deploying speech
  enabled applications in network environments.
•  

7
In the year of 2004 cont.

• Potential collaboration