Research and teaching Computer Vision and Artificial Intelligence

One of the most interesting challenges in modern science and engineering is the building of
intelligent machines which can interact with humans and perform useful tasks autonomously. In
light of this requirement, my research programme focuses on developing computational analogue
for basic human cognition and exploiting the strengths of computers to take full advantage of these
capabilities. Specifically, I am interested in intelligent systems for activity/behaviour analysis and
recognition that brings together artificial intelligence, computer vision, machine learning, insights
from psychology and physiology, algorithms and a great deal of computation.

The main idea behind this research was to investigate methods for building an efficient multimedia
system for document-based automatic indexing and retrieval of multimedia data captured from
multimodal environments such as meetings, conferences, etc. Both empirical image processing,
video segmentation methods and document analysis approaches are studied to bridge the
gap between temporal data and static information. In order to provide useful access points, the
captured audio-visual data were fragmented into reasonable distinct smaller segments using
projected documents, which provide meaningful semantic pointers because they appear at
a specific time, remain in visual focus for a definite duration and summarize the presenter’s
discourse at that time. A Visual Signature-based identification of low-resolution document images
is developed to link original electronic documents with the temporally segmented captured
multimedia data and image-based retrieval of multimedia information. Matching of signature
is based on both sequential as well as multi-level linear and non-linear fusion of various visual
features. The above-mentioned techniques prove the usefulness of documents as an additional

modality and natural interface, in interacting with multimedia data captured from multimodal
environments.