Research We Do

Empirical Aesthetics

Empirical Aesthetics is an enquiry into the underlying factors that contribute to an aesthetic arousal (from music, poetry, dance, sculpture, animated computer displays, web pages, videos) using modern scientific methods like psychophysical behavioral experiments and eye movement studies.

We concentrate on visual arts and paintings in our study of aesthetics. Among all the aesthetic principles involved in a painting, our current focus of work in the lab concentrates on the aspect of "Visual Balance"

Understanding Visual Balance We do basic research on verifying the theoretical writings on this design principle of Balance. We try to answer questions like -

Structural, Physical Properties affecting Balance Using artworks of the famous painter Piet Mondrian, we try to come up with some heuristics as to what structural physical properties of a painting contribute to the overall visual balance of the composition?

Visual Metaphors, Creativity and Cognition

To attempt to understand the structure of metaphor is to attempt to understand the structure of something that contributes to the intelligibility of the world. Metaphors are not simply literary devices, but something active in understanding, perhaps even the very basis of language. Metaphors organize our experience, uniquely express that experience, and create convincing realities.

Visual communication is such a domain where we can find metaphors. This includes various works of art, advertisement, etc.

The initial problem for developing studies of visual metaphor centers around on the various ways researchers define the term. Differences arise in the amount of emphasis given to factors such as -

Our research involves issues like -

Cognitive Modeling of Web-navigation

Several attempts at cognitive modeling of web-navigation behaviour (SNIF-ACT, MESA, CoLiDeS, CoLiDeS+) have focused entirely on the semantic information derived from textual elements and repeatedly ignored the role played by graphical and visual information. While research on visual search has emphasized the efficiency of our visual system at perceiving salient objects, there have been no attempts at including information from the visual modality into cognitive models of web-navigation.

We investigate the role played by semantics from textual and graphical elements on a web-page in locating widgets. We also develop new models that incorporate semantic information coming from pictures on a web-page into CoLiDeS. We compare and contrast the performance of the new model with existing models.

Cognitive Vision

Vision is the most studied and most important of human senses. It is fundamental for our survival in this complex environment. It helps us in navigating through this complex world and in identification of objects even from a distance. It is normally so sure and effortless that we naturally assume that it is indeed effortless. Behind this ease there is intelligence so great that it occupies nearly half of our brain's cortex. Human vision has long fascinated the cognitive researchers. Although Vision has revealed many of its secrets to physicists, neurobiologists, psychologists, and researchers in computer vision, many of its secrets are yet to be unveiled. Our vision is highly sophisticated and our understanding of visual processing is only at its stage of infancy. We even don't know completely how visual signals are carried in our brain.

Vision can be described as a complex information processing task. It is the process of discovering what is present in the world and where it is, from the images at the eyes. Cognitive vision is an interdisciplinary field aimed at understanding this complex task and in developing automated computational models that emulate the tasks carried out by our visual system. The synergy between biological and computer vision research can give substantial insights into the processing carried out in our human visual system.

The modularity of our visual system had given birth to several sub fields in the field of cognitive vision such as those relating to color perception, motion perception, 3D vision etc. Current focus of research in the lab includes -

Music-IR and Cognition

Music-IR and Cognition The focus of our work is to build a recommendation system for Music in general. We work on content based recommendations systems such as raaga recognition, in conjunction with collaborative filtering based methods. Our test bed is Indian classical and to an extent, film music, owing to the well observed raaga-rasa theory in Indian musicology.

Discourse Generation

Current Narrative Generation systems do not concentrate on the order in which the events of a story are presented to the viewer/reader. We wish to develop a discourse module that generates a non-linear narrative that is more engaging and at the same time comprehensive than its simple linear counterpart. In order to generate a narrative that is comprehendable, it is essential to study the processes of Narrative Comprehension from a Cognitive stand point and use the findings as heuristics for developing such a model.

For this purpose, we have studied how viewers' recall and comprehension are affected when the order of presenting the information is varied. We have conducted experiments using a commercial film, that was edited in different ways for experimental purpose. Participants were required to answer questions related to the narrative and write a summary of what they have seen. The results obtained were studied in the framework of the Event-Indexing Model, a Cognitive model for Narrative Comprehension, by Zwaan et. al.