This university seminar welcomes scientists from Carnegie Mellon's Human-Computer Interaction Institute. They will demonstrate and discuss the educational application of two innovative video technologies they are developing, Informedia and CareMedia.

Informedia
Informedia digital video research focuses on improving access to broadcast video information through speech recognition, computer vision, and natural language processing techniques. Automatically derived descriptors for the video are used to construct information visualization interfaces for querying, summarizing, and browsing the video. For over 12 years, Informedia has recorded and analyzed several hours a day of CNN, Chinese, and Arabic news video.

CareMedia
CareMedia research centers on the automatic analysis of audio and video for behavioral research. Their most recent work captured video from 23 cameras in public spaces of a nursing home dementia ward. Video was captured from each camera, 24 hours a day for 25 days. The data collected totaled over 13,000 hours of video stored on 35 Terabytes of hard disks. Clearly, this volume of data precludes manual analysis.

CareMedia's interdisciplinary research is developing, integrating, and refining a suite of tools supporting the automatic collection, annotation, access, analysis, and archiving of such massive amounts of behavioral data. These tools capture a continuous audiovisual record of individual and group activity in various settings and apply machine intelligence technology to automatically process that record for efficient use by analytical observers to monitor situational behavior over time. The annotated record provides a level of completeness not feasible with human observers, and allows, for the first time, large-scale longitudinal clinical and behavioral research based on continuously captured and processed data, enabled through extensible interfaces accessing such voluminous records in a user-friendly utilitarian manner.

Date: February 16, 2006 at 4:00pm
Location: 523 Butler Library