Research
Seungjin Choi's Current Research
My Current Research Interests
-
Probabilistic/Statistical Learning
- Indepdent component analysis (ICA) and principal component
analysis (PCA)
- Kernel machines: Support vector machines (SVM), kernel PCA
- Bayesian learning and graphical models
(including variational Bayes and belief nets)
- Mixture models
- Various learning machines and algorithms
for pattern recognition (face and character recognition)
-
Neural Computation
- Mathematical theory for neural computation
- Unsupervised learning
- Sensory processing and information representation
- Brain-like computation
- Brain computer interface
- Data analysis and visualization
- Information geometry
-
Computational Biology (Bioinformatics)
- One of my very recent interest
- Biostatistics and machine learning for bioinformatics
- DNA microarray data analysis
-
Computational Hearing
- Understanding multiple
simultaneous speeches: Cocktail party speech recognition
- Computational auditory scence analysis (CASA)
- Natural sound analysis
-
Computational Vision
- Understanding and development of human vision-like system
- Natural scene analysis
- Learning manifolds for perception
- Related with probabilistic learning and neural computation
-
Signal/Image Processing
- Blind signal processing and higher-order statistics
(multichannel blind deconvolutio and source separation)
- Medical imaging (especially, brain imaging such as
PET and fMRI)
- Adaptive algorithms (including natural gradient)
- Multisensory processing
- Brain signal processing (especially biomedical signal
analysis such as EEG and MEG)
- Image/Speech de-noising
- Wireless communication signal processing
Current Projects
-
Development and VLSI implementation of ICA methods
for efficient source separation
- Supported by Korea Ministry of Science and Technology
- Septermber 2000 - August 2003 (3 years)
- Collaboration with KAIST (Prof. S.-Y. Lee) and
RIKEN Brain Science Institute (Profs. S. Amari and A. Cichocki)
-
Comparative study of several models for segregation of cocktail
party speech, based on spatial localization of speakers
- Supported by KOSEF
- September 2001 - August 2003 (2 years)
- Collaboration with ICP, INPG, France (Drs. F. Berthommier
and H. Glotin)
-
Methods of efficient feature extraction for EEG analysis
- Supported by ETRI
- June 2001 - November 2001 (6 months)
-
Development and application of myocardial blood flow
quatification method using dynamic PET
- Supported by KOSEF
- September 2000 - August 2005 (5 years)
- Collaboration with Seoul National University, Medical
School (Prof. D.S. Lee)
-
Development of artifical vision and hearing systems
- Supported by Korea Ministry of Science and Technology
- November 1998 - July 2001
- One of Braintech 21 national project
-
Co-work with Brain Science Institute, RIKEN, Japan
- Supported by Brain Science Institute, RIKEN, Japan
- Collaboration with Advanced Brain Signal Processing Lab
(directed by Prof. A. Cichocki),
Brain-style Information Systems
Research Group (directed by Prof. S. Amari)
- Source separation methods for biomedical signal analysis
- Speech segregation
- Natural gradient learning