How AI Can Help Robots Reduce Hand Tremors

Neurological movement disorders can severely hamper the quality of life for sufferers around the world.  New research from New York University explores how artificial intelligence and robotics can help.

Their model, which they’re making available for others to deploy and build upon, aims to provide support to the one million people around the world who have Parkinson’s disease, which is just one of the conditions that can result in hand tremors.

Traditionally, technology to solve this problem has been in the form of complex wearable exoskeleton suits or neurorehabilitative robots.  For such devices to work, however, they need to be able to precisely predict the involuntary movements made by the wearer in real-time.  A lag of even 10 or 20 milliseconds can prevent effective compensatory motions being made by the machine.

Artificial assistance

This is where AI comes into play.  The researchers utilized a huge dataset that had been collected by the London (Ontario) Movement Disorders Centre to train a machine learning model, referred to as PHTNet (or Pathological Hand Tremors using Recurrent Neural Networks to give it its full title).  They augmented this with an array of small sensors to enable them to analyze the hand movements of 81 patients, with the neural network then better able to predict the movement of each individual.

When the system was put to the test on nearly 25,000 different samples, it was able to achieve a confidence rate of approximately 95%, which the team believe renders it ready for market.

“Our model is already at the ready-to-use stage, available to neurologists, researchers, and assistive technology developers,” they say. “It requires substantial computational power, so we plan to develop a low-power, cloud-computing approach that will allow wearable robots and exoskeletons to operate in patients’ homes. We also hope to develop models that require less computational power and add other biological factors to the inputs.”

Facebooktwitterredditpinterestlinkedinmail