| EPSRC Reference: |
GR/H87629/01 |
| Title: |
VIDEO MAIL RETRIEVAL USING VOICE IED4/1/5804 |
| Principal Investigator: |
Young, Professor SJ |
| Other Investigators: |
|
| Researcher Co-Investigators: |
|
| Project Partners: |
|
| Department: |
Engineering |
| Organisation: |
University of Cambridge |
| Scheme: |
Standard Research (Pre-FEC) |
| Starts: |
01 April 1993 |
Ends: |
30 September 1996 |
Value (£): |
301,667
|
| EPSRC Research Topic Classifications: |
| Human Communication in ICT |
|
|
| EPSRC Industrial Sector Classifications: |
|
| Related Grants: |
|
| Panel History: |
|
|
Summary on Grant Application Form |
|
To develop robust unrestricted keyword spotting algorithms for use in video document retrieval To adapt existing text-based information retrieval techniques to work effectively on voice data To develop and demonstrate a practical system providing video document retrieval using voiceProgress:The project is organised in 3 stages, each lasting one year and each culminating in a prototype demonstration system. The first stage prototype was completed in September 1994 and performed successfully with predefined keywords operating on messages from known speakers. It was also supplied with a specially tailored interface. Current work on the second stage prototype will extend this to allow unknown speakers. The retrieval strategies used will also be enhanced to improve weighting of query search terms by incorporating further techniques adapted from text-based information retrieval. The research which underpins the above system development is focused on improving word spotting and developing information retrieval to work effectively on voice data. To facilitate experimentation and assessment, a database of spoken messages was recorded and orthographically transcribed. The messages cover ten categories chosen to reflect the anticipated needs of local users, and a set of 35 keywords were chosen to cover them. Using this database, word spotting algorithms based on whole-word and sub-word Hidden Markov Models (HMMs) have been developed and a basic retrieval engine using inverse document frequency weighting has been produced. For the speaker dependent case, retrieval using voice is around 90% of that which can be achieved on the same messages using text. However, speaker independent operation is rather worse and speaker adaptation techniques are being developed to remedy this. Overall, the work is proceeding well and on schedule. A number of reports, and conference and workshop papers have been presented or accepted and due deliverables have been completed. For further details see WWW page:http://svr-www.eng.cam.ac.uk/Research/Projects/video_mail_retrieval_voice.html
|
|
Key Findings |
|
This information can now be found on Gateway to Research (GtR) http://gtr.rcuk.ac.uk
|
|
Potential use in non-academic contexts |
|
This information can now be found on Gateway to Research (GtR) http://gtr.rcuk.ac.uk
|
|
Impacts |
|
| Description |
This information can now be found on Gateway to Research (GtR) http://gtr.rcuk.ac.uk |
| Summary |
|
| Date Materialised |
|
|
|
|
Sectors submitted by the Researcher |
|
This information can now be found on Gateway to Research (GtR) http://gtr.rcuk.ac.uk
|
| Project URL: |
|
| Further Information: |
|
| Organisation Website: |
http://www.cam.ac.uk |