Artificial Intelligence Can Help Control Fingers Of Bionic Hand Using Natural Commands

An AI-based prototype interface offers users a chance to control their prosthetic arm using natural commands. 

Motorica’s hand and forearm prosthetic has been designed to bring a “natural” control to users.

Source : Motorica

Bionic arms have proven to be indispensable to those in need over the years. However, most modern prosthetic control systems are capable of a limited number of functions, including the likes of clenching, flexion/extensions, and left/right rotations. The reason behind this is primarily because the use of gestures in bionic prostheses requires presets of instructions, and then too, there’s a general lack of decision-driven support systems that can enable natural control of each finger of prosthetics.

Artificial intelligence (AI) can help rectify that issue.

An AI-based prototype interface developed by Moscow-based Motorica, which also has an office in Bengaluru, offers users a chance to control their prosthetic arm using natural commands. 

Motorica’s hand and forearm prosthetic has been designed to bring a “natural” control to users, aided by a neural network that remains a couple of steps ahead of the users to smoothen the process even further. 

If you have played shooting games on consoles and used aim assist, you may have an idea of how the new prosthetics technology functions. For those unaware, aim assist in video games allow players to lock on to an in-game target as soon as they can take their gun’s crosshair near the enemy. This comes in handy because trying to aim using console controllers is difficult and slower when compared to gaming with a PC mouse. So, the game’s engine uses aim assist to quickly help players lock on to an enemy as long as they target the crosshair in a general area closer to the target.

Similarly, Motorica’s new sensors accurately transmit movement of each muscle and tendon to an AI-driven decision support system. This helps predict and control the movement of each finger of a prosthetic hand to suit the commands of the user, to bring the experience closer to natural movements. 

The AI-based approach reduces the response rate to 0.02 seconds (from the moment a command is received), far exceeding the speed of most electromyography-based control systems. 

Motorica has tested its new technology on nine test subjects, and plans to train its neural network on more users in Bengaluru in 2023. This data will help create a database of signals for the AI to learn from, speeding up the study process for each new user. 

“The algorithm we have developed is to bring the hand movement closer to the natural one. At the same time, in some cases the neural network will be a couple of steps ahead of you,” said Motorica CBDO Vasilii Khelbnikov. “Now, the main challenge is to collect a large amount of data. Our vision of future prosthetics control is a combination of AI prediction and sensitive feedback.”