Technical Reports (CIS)

Document Type

Thesis or dissertation

Date of this Version

1-1-2001

Comments

University of Pennsylvania Department of Computer and Information Science Technical Report No. MS-CIS-01-24.

Abstract

Humans use gestures in most communicative acts. How are these gestures initiated and performed? What kinds of communicative roles do they play and what kinds of meanings do they convey? How do listeners extract and understand these meanings? Will it be possible to build computerized communicating agents that can extract and understand the meanings and accordingly simulate and display expressive gestures on the computer in such a way that they can be effective conversational partners? All these questions are easy to ask, but far more difficult to answer. In this thesis we try to address these questions regarding the synthesis and acquisition of communicative gestures.

Our approach to gesture is based on the principles of movement observation science, specifically Laban Movement Analysis (LMA) and its Effort and Shape components. LMA, developed in the dance community over the past seventy years, is an effective method for observing, describing, notating, and interpreting human movement to enhance communication and expression in everyday and professional life. Its Effort and Shape component provide us with a comprehensive and valuable set of parameters to characterize gesture formation. The computational model (the EMOTE system) we have built offers power and flexibility to procedurally synthesize gestures based on predefined key pose and time information plus Effort and Shape qualities.

To provide real quantitative foundations for a complete communicative gesture model, we have built a computational framework where the observable characteristics of gestures - not only key pose and timing but also the underlying motion qualitites - can be extracted from live performance, either in 3D motion capture data or in 2D video data, and correlated with observations validated by LMA notators. Experiments of this sort have not been conducted before and should be of interest not only to the computer animation and computer vision community but would be a powerful and valuable methodological tool for creating personalized, communicating agents.

Share

COinS
 

Date Posted: 06 November 2006