Rice University scientists have created a system by which mobile users can quickly determine their location indoors without communicating with the cloud, networks or other devices.

The battery-saving scheme uses image recognition and “hashing,” a method that reduces key details in a photo to short strings of numbers. To determine a location, the system hashes a photo from the user’s camera and compares it against a pre-downloaded, highly compressed location database.

Capsule allows users to find their location quickly and securely in an indoor venue, such as a mall. Image credit: Public Domain Pictures. “The core of our system is a hashing-based image-matching algorithm that is more than 500 times cheaper—both in terms of energy and computational overhead—than state-of-the-art image-matching techniques," says Anshumali Shrivastava, assistant professor of computer science.

Describing how the Camera-Based Positioning System Using Learning (Capsule) system might be used, Shrivastava cites the example of a shopping mall. The mall owner would need a gallery of images of the mall's interior, and Capsule would scan those images for key features, such as store marques, escalators and kiosks. Rather than storing the images, the system stores a table of hashes that serve as lightweight image fingerprints and can be computed quickly.

In tests on a commercially available smartphone, Capsule calculated locations in less than two seconds with greater than 92% accuracy using under 4 joules of energy. According to Shrivastava, the proof-of-concept application uses a combination of machine learning and "inexact computing" to address three of the primary problems facing mobile application designers.

“Privacy, computations and energy are the big challenges,” he says. “Inexact computing helps with all three. In short, it allows us to determine answers with something less than 100% confidence."

According to Shrivastava, there are many situations in which a miniscule loss of confidence—say, 1% or less—is inconsequential with regard to the ability to provide a good measurement of a user's location. Yet that tiny difference in accuracy can provide considerable savings in the number of computations and the amount of energy required to perform a query.

For example, a traditional image-matching technique that Shrivastava and colleagues used for comparison with Capsule consumed over 500 times the energy and took almost 17 minutes to complete a single location query when computations were performed on a smartphone. For that extra energy and time, the accuracy improved to 93.4%—less than 2 percentage points better than Capsule's accuracy.

“Cloud-based machine-learning applications are getting a great deal of attention, but cloud-based solutions have inherent privacy drawbacks and they are typically computationally and energy- intensive,” Shrivastava says. “Capsule shows that a ‘cloudless,’ probabilistic approach can be a viable and more sustainable alternative.”