Designing Robust Multimodal Systems for Diverse Users and Mobile Environments
Speaker: Sharon Oviatt , Oregon Graduate Institute
Date: August 30 2001
Time: 4:00pm
Location: NE43-941
The advent of multimodal interfaces based on recognition of human speech, gaze, gesture, and other natural behavior represents just the beginning of a progression toward pervasive computational interfaces that are capable of human-like sensory perception. Such interfaces eventually will interpret continuous simultaneous input from many different visual, auditory, and tactile input modes, which will be recognized as users engage in everyday activities. They also will track and incorporate information from multiple sensors on the user's interface and surrounding physical environment in order to support intelligent adaptation to the user, task and usage environment. In the future, these adaptive multimodal-multisensor interfaces have the potential to support new functionality, to achieve unparalleled robustness, and to perform flexibly as a multifunctional and personalized mobile system.
In this talk, I will focus on how current multimodal systems can be designed for robustness. In two recent studies involving over 4,600 multimodal commands, a multimodal architecture supported error suppression ranging between 19 and 41%. Improved robustness also was greater for "challenging" user groups (accented vs. native speakers) and usage contexts (mobile vs. stationary use). These results indicate that a well-designed multimodal system not only can perform more robustly than a unimodal system, but also in a more stable way across varied real-world users and usage contexts. Largely for these reasons, during the next decade we are increasingly likely to see promising but error-prone new input modes embedded within multimodal architectures in order to harness and stabilize them more effectively.
Sharon Oviatt is a Professor and Co-Director of the Center for Human-Computer Communication (CHCC) in the Dept. of Computer Science at the Oregon Graduate Institute of Science & Technology (OGI). She previously has taught and conducted research at the Artificial Intelligence Center at SRI International, and the Universities of Illinois, California, and Oregon State. Her research focuses on human-computer interaction, spoken language and multimodal interfaces, and mobile and highly interactive systems. Examples of recent work involve the development of novel design concepts for multimodal and mobile interfaces, robust interfaces for real-world field environments and diverse users (children, accented speakers), and conversational interfaces with animated software "partners." This work is funded by grants and contracts from the National Science Foundation, DARPA, ONR, and corporate sources such as Intel, Motorola, Microsoft and Boeing. She is an active member of the international HCI and speech communities, has published over 70 scientific articles, and has served on numerous government advisory panels, editorial boards, and program committees. Her work is featured in recent special issues of Communications for the ACM, Human Computer Interaction, and IEEE Multimedia. Further information about Dr. Oviatt and CHCC is available at http://www.cse.ogi.edu/CHCC.
See other events that are part of AI Colloquium Series Spring 2001
See other events happening in August 2001
