Catalyst for the Digital Home: Intel’s Fourth-Generation UI Research
Part 2 of a Series

Intel’s Digital Home Innovations team, working with leading university researchers and visionary interface designers, is advancing the UI technology to provide consumers with a new dimension of control. The effort points the way to future interfaces that will harness the performance of consumer electronics platforms based on Intel technology.

In Part 1 of this series, Gary Palangian and Randy Dunton of Intel’s Digital Home Innovations Team discuss the evolution of CE user interface. In this article, they describe Intel’s ongoing UI R&D program, and share some of the results to date.

Q. Why is Intel involved in UI research and development?
We believe that if you get the user interface right, the popularity of digital home platforms will take-off, as consumers discover how easy it can be to access the new content and services supported by the performance and headroom of the system.

Remember the explosion of the PC adoption once Apple created the first consumer PC with a GUI. Suddenly the barrier of entry for everyday users dropped, and the price of the system was affordable as well.

The cost for a digital home media center device is coming down, and our next-generation silicon will reduce it even further. So what we need is the right UI. Naturally, a user interface with great consumer appeal may actually require a very complex set of systems under the hood.

Q. Will human interface design is about help users manage this complexity?
Meeting the user interface challenge involves research in Information Architecture and Human Factors Engineering, a field that originated in efforts to redesign aircraft cockpits that saved pilots from being overloaded with control choices. Home viewers using today’s digital home products can confront a similar set of problems.

Think for a moment about how far the TV control environment has evolved. From the first TV with just a few simple knobs, to a single remote, to a VCR which added a second remote, to today’s multi-device home theatre installations which can have anywhere from three to six remotes!

We will see even more complexity as the digital home keeps evolving from a LAN-based environment to widespread Internet connectivity. Bear in mind that if consumers feel overwhelmed by complexity, many of the most exciting usage models may not get off the ground.

Our role is to invent new ways to make potentially complex choices effortless, intuitive and engaging, even for non-technical consumers. Our current UI uses only two buttons, and we think that in the end we may only need four.

Q. How are you optimizing the human-machine interface?
The building blocks include graphical elements, input devices, and information. The graphical user interface must transfer much more information at a glance. We are looking at 3D graphical user interfaces that use three-dimensional techniques to aid navigation and map media representations onto 3D objects, so people can easily visualize the available content and explore the available content before to making a choice.

Our brains are designed to move in space, to search, explore and manipulate in a three-dimensional world. A 3D graphical UI design can lead to a much more intuitive experience than any of us have ever experienced outside of the real world. This will truly make the interface an immersive experience.

Q. What about human interface devices?
The interface must also increase the ability of the viewer to make choices through new input devices. The conventional remote, with its up/down, left/right and select button, has not evolved much since the first remotes appeared almost 50 years ago. To enable an enhanced degree of control from the 10-foot viewing distance, we are looking at free-form pointing devices, such as gyro-remotes as well as limited amounts of voice input to aid in search.

And finally the information, or metadata, we provide the viewer to make choices or recommendations must be of the highest quality. Metadata comes from various sources, such as online databases, or it can be machine-generated.

Q. Who has worked with Intel on this project?

Fourth-generation CE interface with 3D graphics (Source: Schematic)

In addition to a larger spatial navigation framework and an aesthetically pleasing design created for us by Schematic, the GUI has an integrated search and browse capability. This new information architecture was discovered at the Association for Computing Machinery (ACM) Special Interest Group on Computer-Human Interaction (SIGCHI) conference in 2006, and it is based on years of work by a professor in the School of Information at UC Berkeley, Marti Hearst, who has been researching new ways to enable fast searching of vast libraries of information.

Q. What have you learned so far?
In the summer of 2006 we created our first working prototype, and we learned a number of surprising things. The most important discovery was that our UI became sluggish very quickly due to the large system loads it created.

We fixed many of these issues by creating new and powerful databases tailored to feed the UI, we have re-architected the entire foundation of the UI to deal with the collection of metadata, and connect the local database to external databases.

We also explored ways to create metadata by analyzing personal content, such as photos and home movies. Examples of metadata tags would be things like sunsets, people and trees – labels that people can relate to. This area has yielded very good results, such as the ability to find pictures in a vast picture library, but it also results in large system loads. The good news is that thanks to Moore’s Law, this problem can be solved by the continuing development of low cost, high performance processors.

The UI is controlled using a gyroscopic pointing device, also known as a free-form pointer, which is usable from a 10-foot distance. This device is based on available technology created by Gyration*. A big part of our current work involves integration of all of these elements into platform solutions.

Q. How important is metadata in future platforms?
Once you begin to design a new UI with a search and browse engine for digital content, you immediately realize there is not a great deal of high quality and available metadata. You can get metadata from limited commercial sources, but personal photos and movies do not have enough metadata to be meaningful for a new UI. At Intel we are researching algorithms capable of performing content-based media searches. For example, you could search for photos or video with trees, snow, or lakes in them.

The analyzer that interprets the content on the digital image must be trained, and we have attained hit-rates of 90 percent. The work on adding metadata to personal content is based on some work we have conducted in the Intel Research Lab at the University of Pittsburg. The “Diamond” project involves interactive search of terabyte-scale non-indexed collections of complex data, such as photo collections. We have adopted it to the digital home environment, and added modifications to increase performance.

Q. Will our future CE devices do the searching for us?
One of the most important parts of a fourth-generation UI is that a person needs recommendations to be able to find content that they may otherwise never know about. The first step is for the system to understand the person or people it is serving, through biometrics or voice-print analysis, or requiring the user to sign-on.

Once it recognizes who you are, the inference engine can make recommendations based on what it has learned about your prior usage patterns. Because the inference engine has information about future media broadcasts, or perhaps the entire database of every movie ever made, it can make relevant suggestions. This provides a level of personalization that moves beyond direct searching and browsing to a more intelligent system that can show me something that is appeals to me.

The system will also reduce the need to search in the first place, by organizing information in a way that is more meaningful for you personally.

Of course having adequate processing headroom in the platform is essential for making all of this possible. We are running the prototype interfaces on PCs today. Ultimately these applications will migrate to high-performance embedded Intel® architecture platforms.

Q. What’s next for your team?
Our first objective in 2006 was to develop a prototype system, running on a PC platform, for demonstration and usability studies.

For 2007 we are re-architecting the foundation of the system to better handle the load generated by the UI. To make the prototype closer to the real end-product, we have migrated our code base to Linux.

We’ve identified several software technologies supporting the user experience that are well suited to hardware acceleration through multi-core and vector processing. We need to prioritize the most promising opportunities to allow competitive differentiation and planning for market deployment.

Our early work in inference engines indicates more work will required to make it usable. In the area of creating metadata for personal content we are now evaluating how well we can recognize and identify people, matching a name to a face. We are also exploring data visualizations that help in the process of looking through the result sets generated from an interactive search or browse of a large library. Finally, we are preparing ourselves to have a working UI to begin usability research involving multiple demographics.

As odd as it may sound, aesthetics plays a large role in a GUI. Good looking things work better. A recent paper in SIGCHI pointed out that, all else being equal, a more aesthetically pleasing GUI scored better on customer satisfaction and task completion. So we plan to give this area of the UI some further work.

It is important to note that next-generation user interfaces are about more than just cool design. A great UI must organize vast amounts of information, while keeping the user experience engaging and effective, while also providing emotional impact and a sense of attachment to the product.

User interfaces will take advantage of the performance of embedded Intel architecture and graphics acceleration in our next-generation silicon offerings specifically designed for CE platforms. We believe that these silicon building blocks, combined with our research, will become the catalyst to enable a digital home platform which will delight and entertain a large segment of the population, and thanks to Moore’s law at an affordable cost.