K-Glass smart glasses mimic the human brain to improve energy efficiency

 

By Nick Lavars February 19, 2014

In basing the artificial neural network on the brain's central nervous system, the team says it was able to compartmentalize the processing of data, resulting in less congestion and significantly improved energy efficiency

It took a heavyweight like Google to bring the notion of head-mounted devices to the mainstream, but other developers are also testing the waters and pushing the boundaries of what's possible to achieve in the smart glasses space. Exhibit A is K-Glass, a wearable, hands-free display developed by researchers at the Korea Advanced Institute of Science and Technology (KAIST).

Much like its counterparts, K-Glass is designed to offer users an everyday augmented reality (AR) experience. According to the developers, users will be able to walk up to a restaurant and have its name, menu, available tables and a 3D image of different food displayed in front of their eyes.

A point of difference that could distinguish the K-Glass technology from other head-mounted displays, and one emphasized by the researchers, is the approach used to generate the augmented reality experience. Rather than using methods such as algorithms, facial recognition, motion tracking, barcodes and QR codes to establish and deliver a virtual reality like other head-mounted displays, K-Glass is designed to replicate the process our brains use to establish our surroundings.

This all revolves around an AR processor based on the Visual Attention Model (VAM), which reproduces the ability of the human brain to categorize relevant and irrelevant visual data.

When we process visual data we use sets of neurons that, though connected, work independently on different stages of the decision making. One set of neurons completes part of the process and relays the information onto a the next set, before ultimately a set of decider neurons determine what data is required and what can be done away with.

In basing the artificial neural network on the brain's central nervous system, the team says it was able to compartmentalize the processing of data, resulting in less congestion and significantly improved energy efficiency. According to its creators, K-Glass can deliver 1.22 TOPS (tera-operations per second) while running at 250 MHz, using 778 mW powered by a 1.2V supply. The team says this equates to a 76 percent improvement on power consumption of similar devices.

"Our processor can work for long hours without sacrificing K-Glass's high performance, an ideal mobile gadget or wearable computer, which users can wear for almost the whole day," says Hoi-Jun Yoo, Professor of Electrical Engineering at KAIST.

The team presented its research this month at the International Solid-State Circuits Conference (ISSCC) in San Francisco, California.