A Deep Dive Into Gesture Recognition

Whether it is for business or pleasure, scuba diving carries many risks with it. From equipment malfunctions to underwater currents and marine life encounters, divers face a variety of potential dangers beneath the waves. To mitigate these risks, divers often venture underwater in groups for added safety. The buddy system has become a fundamental principle in diving, where each diver is paired with another to watch out for each other.

However, even with a buddy system in place, underwater communication can be challenging due to limited visibility and the inability to communicate verbally. This presents a significant obstacle if one diver notices trouble but struggles to get the attention of their buddy or effectively communicate what they have observed. In such situations, swift and decisive action becomes crucial to ensuring the safety of all divers involved.

Exactly how to react in these situations is a subject that is still open to debate. One potential solution was recently proposed by a team of researchers at Qingdao University. They have developed a waterproof electronic glove that is capable of recognizing a number of hand gestures. By interpreting these gestures, wireless signals can be transmitted to other divers to alert them when another member of the group is attempting to communicate, even if they are not looking at them.

The underwater sensing system was inspired by the tube-like feet of a starfish. It consists of an array of tiny pillars etched into a sheet of waterproof plastic called polydimethylsiloxane. The pillars are then coated in a conductive layer of silver. A pair of these sheets are pressed together, with the conductive pillars on the inside. After sealing them together, a watertight bond is formed.

When these sensors are flexed, the deformation in the material can be detected by the electric signals they produce. It was demonstrated that anything from the light touch of a piece of paper laying on top of the sensor to a hard press could be detected. Ten of these sensors were attached to the knuckles and first finger joints of a glove to complete the hardware design of the system.

Knowing that a particular sensor has been flexed in some way and recognizing that a specific gesture is being performed are two very different things. So, in order to accurately recognize complex gestures, the researchers built and trained a machine learning model. Initially, the system was trained to recognize a set of sixteen hand signs that represent words or phrases. The glove was then tested both on land and underwater, and it was found to be highly accurate. The correct hand gesture was recognized in 99.8 percent of cases, on average. It also was shown to be capable of enabling real-time communication, with response times of around 23 milliseconds.

Looking ahead this glove could greatly enhance underwater communication by providing a simple, inexpensive, and accurate solution. Given these attributes, the researchers believe that the technology holds tremendous potential for a wide range of applications in the field of underwater communications.