Lip-Motion/Voice Recognition Videos
Description
This consists of 20 randomly selected individuals who were videotaped reciting the FitnessGram Pacer Test script and an arbitrary sentence in two separate videos respectively. Each person has two videos in each of their folder, which has been labeled with numbers to preserve the privacy of their names. Originally, this data was collected to train/test a compound biometric system that consists of lip-motion and voice authentication. The training videos averaged between 30 to 45 seconds and the testing videos averaged between 5 to 8 seconds each.
Files
Steps to reproduce
A camcorder is set up roughly 5 feet away from the person and positioned to record from the neckline, up, with a white LED photographic light next to the camcorder set on a medium brightness to illuminate their face. A white backdrop is used to eliminate any unwanted background images and white LED photographic lights on opposite side of the backdrop to wash out any shadows cast by the person.