Google’s Project Soli radar is sensitive enough to count sheets of paper and read Lego bricks

By James Vincent

Project Soli uses miniature radar to detect gestures.
Credit: Google

What does the computer interface of the future look like? One bet from Google is that it will involve invisible interfaces you can tweak and twiddle in mid-air. This is what the company is exploring via Project Soli, an experimental hardware program which uses miniature radar to detect movement, and which recently won approval from the FCC for further study.

Imagining exactly how this tech will be put to use is tricky, but a group of researchers from the University of St Andrews in Scotland are exploring its limitations. In a paper published last month, they show how Project Soli hardware can be used for a range of precise sensing tasks. These including counting the number of playing cards in a deck, measuring compass orientation, and even discerning the specific configuration of a stack of Lego bricks.

All this is done using the delicate radar readings from Google’s hardware, which the researchers incorporate into a system they call RadarCat. As with radar used to detect aircraft, the sensors fires harmless electromagnetic pulses at a target object, some of which are bounced back. The way these pulses return varies based on different factors including distance, the object’s density, thickness, shape, and its surface properties.

Combined, these elements create a unique radar fingerprint for each item. The team’s previous work, published in 2016, showed how this approach could be used to recognize different objects, and their latest work — a technique they are calling Solinteraction — goes even further to produce more nuanced sensing results.

Speaking to The Verge, researcher Hui-Shyong Yeo said the new technique enables all sorts of additional use cases. “The sensing technique remains rather similar [but] the main contribution is the vast exploration into [...] the counting, ordering, stacking, movement, and orientation of different objects, such as cards and Lego blocks.”

The researchers suggest a number of applications, for education, entertainment, and more. Soli radars could be embedded in tabletops to recognize pieces in board games, for example, or built into smart homes to monitor the presence (or absence) of certain items.

Although these tasks can be achieved with other methods, using radar has unique advantages. For example, unlike with RFID chips, there’s no need to alter the object you want to monitor. And unlike with object recognition that uses cameras, there are no privacy concerns, or need to maintain line-of-sight or good lighting conditions. Radar works through different materials and in the dark as well as the light.

There are limitations, of course, including some caused by the high sensitivity of the radar readings themselves. Their delicacy means that small changes to the target object can make it unrecognizable. For example, the researchers found that when counting playing cards, the radar would misclassify the card if it was slightly bent.

It’s not clear what use-cases Google will pursue with Project Soli, but research like this shows there is a huge number of avenues to explore.