Google shows what Soli radar ‘sees’ in Pixel 4 with gesture control

Spread the love

The Pixel 4’s gesture controls don’t seem much more special than many existing implementations at first glance. But looks are more deceiving than you think. Google explains in a new blog post in detail how it has built the technology behind the phone.

The gesture control of the Pixel 4 was immediately positive during our review. Google uses a proprietary technology called Soli for that control. In the Pixel 4, that practically means, for example, that the phone automatically recognizes when a user moves their hand towards the device. Google already announced that the technique worked via radar technology, but many details were not known about it until now. Google has been working on the Soli project since 2015.

In a new blog post, Google describes exactly what the Soli radar sees and how it processes that signal into gesture control. Unlike traditional radar systems, Soli detects movement, not just objects themselves. The Soli chip transmits a modulated 60GHz radio signal, which can then detect displacements of less than a millimeter. The radar then takes into account the Doppler effect to determine the speed of a movement. The distance and speed of a body part registered by the radar can be converted into a three-dimensional image. In a gif, Google shows what the radar ‘sees’. It involves a person walking towards the camera, a hand reaching for the device, and a swipe across the screen.

In addition, the telephone must be able to distinguish between when a gesture is consciously made to operate the telephone and when it is just a random movement. The company did that with machine learning. Google built its own algorithms for this in which thousands of volunteers made millions of hand gestures. That data was combined with hundreds of hours of radar images recording background movements to train a neural network via TensorFlow. This all happened in a low power mode, to make sure that the Soli radar in the phone would not require too much processing power and battery life.

You might also like