Team 27

Cerebelarm: Machine Learning Enabled Vision Assisted Pediatric Prosthetic

Team Members: Danielle Carpenter
Daniel Brennen Martin
Dylan Mitchell
Justin Pettit
Juan Pablo Robayo

Team Mentor: Marco Santello, PhD - SBHSE
Phil Stevens - Director Hanger Clinic

YouTube Link: View the video link below before joining the zoom meeting

Zoom Link: https://asu.zoom.us/j/82270227264

Abstract

The Cerebelarm team aims to produce a pediatric, upper limb prosthetic made through the use of additive manufacturing. This prosthetic utilizes an integrated optical sensor that provides object recognition to dynamically change the prosthetic’s grip based on the object in view. This design concept tackles the difficulty of adjusting grip types within prosthetics seen on the market today, while creating a prosthetic that will hopefully normalize the lifestyle of users at a young age. During this phase of product design, the design team has used various analytical models to verify our design concept. The current design follows engineering guidelines and practices seen within the field. Using the models for validation the team has created designed experiments for our dominant concept to test multiple factors that influence the functionality of the device. The resulting objective functions seek to maximize the accuracy of the machine learning algorithms and minimize the signal to noise ratio of the EMG signal. Prototyping efforts have been primarily focused on the software side of our design. The team has implemented several iterations on visual recognition focused machine learning software using Python’s TensorFlow library. The team has put together a custom database of images based on a standard protocol for data collection, as stock images are insufficient. Most recently, the team is making progress on improving the accuracy of the visual recognition models. The current stage of prototype development is fruitful. The Cerebelarm is making good progress towards meeting the identified customer needs.