Continuous Discriminant Latent Topic Model (CD-LTM) for Facial Expression Recognition and Generation


Grant Data
Project Title
Continuous Discriminant Latent Topic Model (CD-LTM) for Facial Expression Recognition and Generation
Principal Investigator
Dr Chan, Kwok Ping   (Principal investigator)
Duration
42
Start Date
2013-01-01
Completion Date
2016-06-30
Amount
667000
Conference Title
Presentation Title
Keywords
Latent Tropic Models, Facial Expression Recognition, Facial Expression Generation, Latent Dirichlet Allocation
Discipline
Artificial Intelligence (Obsolete),Others - Computing Science and Information Technology
Panel
Engineering (E)
Sponsor
RGC General Research Fund (GRF)
HKU Project Code
HKU 710412E
Grant Type
General Research Fund (GRF)
Funding Year
2012/2013
Status
On-going
Objectives
1) Based on our Temporal Latent Topic Model, we will model the correlation between topics, so that the model is more accurate; 2) Maximum Margin Learning will be incorporated to provide better discriminant information for the model; 3) A Continuous input model will be developed, which should provide better performance as compared to discrete input models; 4) More facial expression data will be collected as the current CMU database used in our study is limited; 5) Facial expression generation system will be built based on the generative Latent Topic Model obtained in the recognition process.