Filming events such as academic conferences and seminars, press briefings, and the defence of graduation papers and dissertations requires considerable effort. Very often, raw footage obtained by the camera operator needs hours of processing, such as editing, before publication, let alone the added challenge of capturing the talk and the slide presentation in parallel.
Denis Korolev, Associate Professor at the HSE MIEM Faculty of Information Technology and Computer Engineering, and Roman Osmolovsky, graduate of the HSE MIEM Department of Information and Communication Technology have created Avtokamera, a solution expected to become a commercial automated video capture device capable of following both the speaker and his or her presentation, while broadcasting, recording, saving, and storing the video.
This video capture solution, which is currently unique, is described in detail in Korolev's paper An Approach to Automation of Video Recording of Spoken Presentations.
Even though the author notes that there is nothing exceptional about the solution at first glance, it certainly merits a more detailed look.
Some DVRs today are capable of automatic synchronous recording on two cameras, but, according to Korolev, these admittedly straightforward and efficient devices lack a microphone input, making the sound quality unacceptable, plus the secondary camera does not provide sufficiently high resolution; therefore, if used to record a lecture, the slide presentation may come out blurred. In addition, the DVRs cannot follow the speaker as he or she moves around.
Korolev combined his expertise in TV production and system and network engineering to create a high-quality, uncompromising technical solution for capturing recurrent events and came up with Avtokamera – a box containing multiple IT-devices, including a video camera to record the event, a Kinect video controller to follow the speaker around, and a computer with a built-in video capture card with two video inputs and corresponding converters to capture both the lecture and the slide presentation, complete with an UPS and a Wi-Fi router.
The software enables the capturing of two video streams and mixing them (or recording them separately upon request), coding for recording, coding for broadcasting, and broadcasting to a specified server; the system is controlled via a tablet or smartphone connected to the device's Wi-Fi network.
"Kinect fills the frontal space with a cloud of dots reflecting light and thus enables the device to track the movements of the person being filmed, instructing the camera to turn in the direction of the person, while preserving the correct frame composition and the subject's flow of movement," explains Korolev. The camera sends the video signal to the computer, where the video is encoded, recorded, and if required, broadcast on the web. "Lectures are often accompanied by slide presentations, but capturing them separately and then combining can be a chore. Therefore, we can also capture the image sent to the projector; once the recording is over, the presentation can be copied to an external storage device or shared via Wi-Fi immediately," explains Korolev.
He notes that Avtokamera is quite expensive at the moment, with a cost comparable to that of a domestically-produced car; however, certain brand-name video conferencing devices are marketed at similar prices while not providing solutions to most of the challenges described.
The developers also admit that they have prioritised reliability over sleek design, focusing instead on placing the device in a sturdy enclosure and making sure it is able to withstand the rigours of everyday use.