Person Identification from Video with Multiple Biometric Cues: Benchmarks for Human and Machine Performance
Abstract
We have compiled a database of images and videos that includes images and videos of approximately 300 human participants. Duplicate sets of images and videos taken from 1 week to 6 months after the first set are available for approximately 200 of these people. The images include 9 standard "mug shot poses". The videos include head rotations, dynamic facial expressions, facial speech clips, and 3 whole-body gait clips. Experiments have been completed comparing the effects of several types of facial motion on face recognition, the effects of face familiarity on recognition from video clips taken at a distance, and on the effects of attention on recognition of moving faces. The results of these studies provide insight into the way motion can facilitate or interfere with the encoding of the invariant face and body features that support recognition.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jan 01, 2003
- Accession Number
- ADA415169
Entities
People
- Alice J. O'toole
Organizations
- University of Texas at Dallas