Don't Just Compress the Video, Compress the Content
Researchers are creating algorithms that remove non-essential information from a video, allowing viewers to watch it faster than in real-time.
In last month’s column, I mentioned I’d been contemplating time compression for, well, some time now. If you’re still with me on this stream-of-consciousness trip around 4D -- and I hope you are -- here’s why: It seems that streaming technologies can be pretty good at folding time, almost as efficient as they have been at image compression and time shifting.
Compression techniques already squeeze relational, spatial, and temporal data from images, enough so that we now discuss H.265/4K combinations in terms of bitrates that were once reserved for D1-quality standard definition.
But what about time compression? Early, clunky time-shifting involved a handful of VHS decks, each one dedicated to recording a specific television series sequentially aggregating the whole series on to a single tape over the course of a few weeks. Today, time shifting is second nature to consumers, with on-demand viewing threatening even linear live television audience numbers.
To truly push the envelope, though, we need faster-than-real-time consumption. In other words, once we’ve figured out how to deliver content to the consumer, in the format of their choice, can we also provide them a standards-based ways to view it 20 to 60 percent faster than in real time?
Applied research into two key areas of “folding time” has been conducted over the past seven years at a number of European research labs, universities, and think tanks. This research focused on both the speed of content delivery and the prioritization of delivered content.
Research revealed it’s possible to speed up delivery of streamed meeting content -- from lectures to all-hands meetings -- without a perceived loss of content.
Researchers then considered whether accelerated content delivery impacts overall fact retention. Initial assessment showed little impact, perhaps due to the fact that shifting the delivery perception is less impactful than actually shifting the speech pattern, an acceleration trick prevalent in U.S. radio ads for auto dealer or credit card offers, where the voice for the credit details shifts markedly from a southern drawl into a “New York minute” speech pace.
Additional research reveals that agenda-driven meetings can also be compressed for both time and content. Think of these two areas of acceleration as the difference between intrafame (M-JPEG) or inter frame (MPEG) compression. In this analogy, agenda-driven meetings offer the chance to set an inter frame (I-frame) and then skip over content that’s not important to the agenda (the equivalent of a P or B frame in MPEG terms).
Researchers noted one major hurdle: Most meetings have an agenda, but the agenda is seldom followed. In most cases, the meeting becomes a sounding board for other issues that need to be addressed by the group, leaving the agenda-based content at less than half of the meeting’s entire “air time” for live participants.
The key, of course, is robust algorithms that split the stream into essential and nonessential content. And, rather than being a hindrance, the fact that lots of meeting time is “wasted” drives home a key point: non-live participants might just be able to sit through the meeting faster than their in-the-room counterparts.
An ideal algorithm removes most non-essential information and assigns it to another video stream, allowing the viewer to choose between delving into the agenda or other areas that were addressed. Trigger points could be set along the way to note where “discussion ensued” on matters not on the original agenda.
Ongoing research into segmenting content now focuses on the use of a written agenda. This agenda could be as simple as an outline, uploaded beforehand, and research continues into the idea of automatically adding keyframe or trigger points to the essential video stream as agenda items are checked off.
So what’s next? That’s up to the industry. Once it decides to focus on folding time and accelerated delivery, hundreds of opportunities are available, for video from the boardroom to the living room.
This article appears in the June 2014 issue of Streaming Media magazine as "Thinking in 4D, Not 4K: Part 2"
Science fiction or science fact? 3D video was a bust with consumers, but 4D could add an exciting new dimension to streaming video.