AI on Adam’s Hand: how does it work?

Adam’s Hand is the World’s First Fully-Adaptive Bionic Hand, that is able to automatically adapt to grasped objects’ shape and size, with no need to select any pre-set grip pattern, as it happens in the case of all of our competitors’ devices.

In the heart of Adam’s Hand there is an algorithm, based on machine learning technology and artificial intelligence, which makes the device use extremely easy and intuitive, thanks to an automatic calibration procedure which takes just a few seconds to be performed.

The origin of Adam’s Hand’s algorithm

The development of our algorithm started in 2016: since the very beginning our team tried to find out the easiest way to turn electric signals gathered from the user’s muscles into command, for obtaining an intuitive and fast control of the prosthesis.

The first versions of the algorithm, to be carried out on a desktop PC, were subjected to a demanding work of optimization, after which the team integrated the algorithm into the electronics board inside the prosthesis.

In 2017 the team presented the first version of the calibration graphic user interface during the Maker Faire in Rome, and reactions collected by the audience demonstrated how intuitive and effective our approach was, allowing to obtain the gesture pattern identification using just two surface electrodes.

Over time, our software has been subjected to the opinion of several field’s experts, as well as users of myoelectric prosthesis, and some of the best Italian orthopaedic technicians. During these tests, the team collected important data and feedback for the improvement of the graphic interface of our App.

The first real test for our algorithm was U.Mano exhibition, organized by Fondazione Golinelli and started in November 2019 in Bologna: Adam’s Hand was showed close to art masterpieces of ancient artists, and visitors had the chance to interact directly with the prosthesis, by controlling it using two electromyographic sensors.

Data extracted from sensors had been processed in order to further improve the control algorithm and they had been made available on our website.

The new Adam’s Hand’s graphic interface

Satisfied with latest results achieved, our team provided the algorithm with a new graphic interface and it realized a new app version for I-pad, which is presented around the world, collecting several feedback during prestigious occasions, such as the CES 2020 in Las Vegas, Nevada, and the acceleration programme at Eilat Tech Centre, in Israel.

The functioning of Adam’s Hand’s software

Once the electrodes are applied on the user’s residual limb, our algorithm immediately starts recording signals of extensor and flexor muscles, displaying them on dynamic graphs.

In this first stage the orthopedic technician checks signals’ intensity and ensures that there is not an excessive overlapping or muscle cross-talk.

Then the automatic calibration starts:

during an assisted-procedure, the algorithm asks the user to contract flexor and extensor muscles, and then to stay at rest, for collecting the muscles’ basal signals. Each contraction takes only 5 seconds and the algorithm automatically selects only the relevant part of the muscular signals.

Through this simple procedure, the algorithm is ready for the pattern recognition of hand opening and closing, although the software graphic interface provides orthopedic technicians with the tools required for small corrections, through simple commands.

Adam’s Hand software will autonomously keep training the algorithm in an unsupervised way, in order to improve the gestures’ identification.

The new version of Adam’s Hand’s App

In the last few months BionIT Labs‘ team has worked on the new version of the App, which will be launched together with Adam’s Hand, and it is able to recognize more than two gestures using only two electrodes, by exploiting a multi-dimensional pattern-recognition approach.

Once the new Adam’s Hand electronic hardware will be released, with the latest version of the algorithm on board, users will be able to perform hand’s opening/closing and wrist rotation at the same time.

Adam’s Hand algorithm and software are just two components of a bigger software platform, through which users and orthopedic technicians will be able to interact and send us feedback about Adam’s Hand functioning, so that we’ll constantly improve the App user-experience.

The results we’re obtaining are moving in this direction: our team’s goal is to create an extremely intuitive device, which can highly improve the quality of life of upper limb amputees.

Skip to content