Speaker
Dr
Daniel DeTone
(University of Michigan)
Description
Communication and collaboration using stored digital media has recently garnered increasing interest in many facets of business, government and education. This is primarily due to improvements in the quality of cameras and the speed of computers. Digital media serves as an effective alternative in the absence of physical interaction between multiple individuals. Video recordings that allow for intimate interaction—the viewer’s ability to discern a presenter’s facial features, lips and hand motions—have been shown to be more effective than videos that do not. To achieve this, a video capture must ensure that the speaker occupies a significant portion of the captured pixels. But camera operators are costly and often do an imperfect job of tracking presenters in unrehearsed situations. This creates the need for a robust, automated system that directs a video camera to follow a presenter as he or she walks anywhere in the front of a lecture hall or large conference room. We present such a system.
The system consists of a commercial, off-the-shelf pan/tilt/zoom (PTZ) color video camera, a necklace of infrared LEDs and a linear photodiode array detector. Electronic output from the photodiode array is processed to generate the location of the LED necklace, which is worn by a human speaker. The computer controls the video camera movements to record video of the speaker. The speaker’s vertical position and depth are assumed to remain relatively constant – the video camera is sent only panning (horizontal) movement commands. The LED necklace is flashed at 70Hz at 50% duty cycle to provide noise-filtering capability. The benefit to using a photodiode array versus a standard video camera is its higher frame rate (4kHz vs. 60Hz). The higher frame rate allows for the filtering of noise infrared such as sunlight and indoor lighting – a capability absent from other tracking technologies. The system has been tested in a large lecture hall, and is shown to be effective.
Primary author
Dr
Daniel DeTone
(University of Michigan)
Co-authors
Dr
Bob Lougheed
(University of Michigan)
Dr
Homer A. Neal
(University of Michigan)
