Ring in the Future

Hands are incredibly versatile tools that serve many purposes in our daily lives. They are capable of gripping and holding onto objects of different sizes and shapes, enabling us to pick up and move things around with ease. Additionally, our hands are capable of recognizing the shape and texture of objects through touch, providing us with crucial sensory information about our surroundings. Our hands also have remarkable dexterity and precision, enabling us to manipulate objects in a variety of ways, such as turning a key or threading a needle. Hands are also integral to communication, allowing us to express ourselves through gestures, sign language, and touch.

Wearable sensors have the potential to enable us to use our hands for more intuitive, and richer, interactions with computers and other electronic devices. Whether it is in gaming, augmented and virtual reality, or ubiquitous computing, the application areas are virtually endless. But to accommodate the many use cases, the number of wearable devices one needs might also be virtually endless. You want to do gesture recognition? Sure, there is a device for that. Object identification, you say? Sure, there is another device for that. How about user identification? You know the drill by now.

A better option may be on the horizon, thanks to the work of a group of engineers at the University of Washington. They have created a wearable device called the Z-Ring that is minimally intrusive, and yet highly versatile. This little ring can sense gestures, and touches, and also offers object and user recognition functionality.

The Z-Ring is worn at the base of one finger on the wearer’s hand, much like an ordinary ring. A single co-located pair of electrodes inside the ring can both transmit an electrical current into the hand, as well as receive the reflected signals. Because the human body is electrically conductive, this approach turns the hand into a sort of antenna. When the pose of the hand changes, or touches other objects, the electromagnetic properties of the antenna system change. To understand what these changes mean, a number of machine learning classifiers were employed by the team.

To handle gesture recognition, a convolutional neural network was trained to recognize the unique electrical signature of five different gestures. Ranging from taps to swipes, these gestures were accurately identified in 93.14% of cases on average in a study with 21 users. In a similar experiment, tangible user interfaces were created from passive objects, like buttons and sliders. A support vector machine classifier was leveraged to decode the user’s interactions with these objects, and the accuracy was quite high. Button presses, for example, were identified correctly 91.8% of the time.

A support vector machine was also put to use for object detection, where it detected six common objects with 94.5% accuracy. In the final trial, a random forest classifier was observed to be 99% accurate in distinguishing between 14 different wearers of the Z-Ring.

This ring might prove to be the jack of all trades that we need for better hand-based input devices, but there is some work to be done yet. The validation studies were conducted with small numbers of users, and also, with small numbers of gestures, passive interface styles, and objects to detect. Without further studies, it is unclear how well Z-Ring would perform under real-world conditions. But perhaps the most significant limitation of the present prototype is that it relies on a commercial vector network analyzer that is far too large to be wearable, and because it has no wireless capabilities, it requires the ring to be physically tethered to a laptop for data transmission. Perhaps with future efforts, the methods laid out in this work will power smaller, more versatile, hand-based input devices.