Design of a Soft Hand Prosthesis For Amputees With a Deep Learning Vision-Based Manipulation System

Presenters

Noah ClarkFollow

Disciplines

Artificial Intelligence and Robotics | Biomechanical Engineering | Biomedical Devices and Instrumentation | Numerical Analysis and Computation | Other Materials Science and Engineering

Abstract (300 words maximum)

Roughly 185,000 amputations occur in the United States yearly, with the number of people living with amputations expected to increase to 3.6 million by the year 2050. To aid amputees with daily activities, researchers use the latest technologies and novel techniques such as myoelectric prostheses. These types of prosthetics are controlled by the electromyographic impulse of the user’s muscles’ nerves. However, control of myoelectric prostheses remains a challenge despite recent technological advances due to overuse injuries and device rejection by the amputee. Furthermore, distinguishing different muscle groups is a cumbersome process which inhibits widespread adoption.

As a solution to these challenges, this research investigates the viability of implementing soft robotics and artificial neural networks to aid amputees with grasping objects. This system would use an embedded camera in the palm of the prosthetic to retrieve an image of the target object. Then, this image will be passed to a convolutional neural network which will identify the target object and determine the suitable grasp type. After a suitable grasp type is determined, the movement of the prosthetic will be controlled using a different neural network. This second neural network will be trained on data collected from finite element method, allowing for finely coordinated, autonomous movement of the prosthetic hand.

The expected outcome of this novel implementation of neural networks and soft robotics is an increased quality of life for people living with amputations by providing them with a prosthetic that will better reflect their desired actions and allow them to fulfill their daily activities. Furthermore, given the significant portion of people living with disabilities in our society, the results of this research is expected to be economically and medically beneficial by providing a basis for the widespread adoption and commercialization of myoelectric prosthetics.

Academic department under which the project should be listed

SPCEET - Engineering Technology

Primary Investigator (PI) Name

Turaj Ashuri

This document is currently not available here.

Share

COinS
 

Design of a Soft Hand Prosthesis For Amputees With a Deep Learning Vision-Based Manipulation System

Roughly 185,000 amputations occur in the United States yearly, with the number of people living with amputations expected to increase to 3.6 million by the year 2050. To aid amputees with daily activities, researchers use the latest technologies and novel techniques such as myoelectric prostheses. These types of prosthetics are controlled by the electromyographic impulse of the user’s muscles’ nerves. However, control of myoelectric prostheses remains a challenge despite recent technological advances due to overuse injuries and device rejection by the amputee. Furthermore, distinguishing different muscle groups is a cumbersome process which inhibits widespread adoption.

As a solution to these challenges, this research investigates the viability of implementing soft robotics and artificial neural networks to aid amputees with grasping objects. This system would use an embedded camera in the palm of the prosthetic to retrieve an image of the target object. Then, this image will be passed to a convolutional neural network which will identify the target object and determine the suitable grasp type. After a suitable grasp type is determined, the movement of the prosthetic will be controlled using a different neural network. This second neural network will be trained on data collected from finite element method, allowing for finely coordinated, autonomous movement of the prosthetic hand.

The expected outcome of this novel implementation of neural networks and soft robotics is an increased quality of life for people living with amputations by providing them with a prosthetic that will better reflect their desired actions and allow them to fulfill their daily activities. Furthermore, given the significant portion of people living with disabilities in our society, the results of this research is expected to be economically and medically beneficial by providing a basis for the widespread adoption and commercialization of myoelectric prosthetics.