PhD Dissertation Defense By Md Abdul Baset Sarker
AI‐Vision on Edge Device: A Next‐Generation Pediatric Prosthetic Hand
Abstract: A major hurdle is that most prostheses, particularly those relying on EEG and EMG
signals, require extensive personalized training. This makes them impractical, if not impossible,
for children to use effectively due to the complex and often frustrating calibration processes.
Previous studies have implemented various vision based prosthesis to reduce the training time.
However, these prosthetic arms were either not portable or lacked adjustability for children.
Furthermore, the implementation of such prostheses on edge devices remains underexplored. To
overcome these limitations, this thesis introduces a novel AI vision-enabled pediatric prosthetic
hand. Our core innovation lies in the strategic integration of AI vision on an edge device,
ensuring real-time control and immediate feedback without requiring individualized training.
This prosthetic hand is 3D-printed using a combination of hard and soft materials. Edge
computing is paramount here; a wrist-mounted micro camera is directly interfaced with a
low-power FPGA for on-device, real-time object detection and precise grasping and gesture
recognition. This local processing minimizes latency, enabling instantaneous actuation in
response to user input. The onboard deep learning (DL)-based object detection and grasp
classification models, performing computations directly at the edge, achieved accuracies of 98%
and 100% respectively. In force prediction, the Mean Absolute Error (MAE) was found to be
0.029. The prosthetic hand features: a) a wrist-mounted micro camera for artificial sensing, b)
real-time object detection, grasp classification, gesture recognition, and distance estimation on
the edge device, and c) low-power operation. This creates a portable prosthetic that enables
pediatric users to achieve accurate grasp control without external processors or calibration.
BIO: Md Abdul Baset Sarker is currently pursuing his PhD in Electrical and Computer Engineering at
Clarkson University, Potsdam, New York. Originally from Bangladesh, he has a background in
Electronics and Communication Engineering and over six years of industrial experience. His research
primarily focuses on Computer Vision and Artificial Intelligence based solutions to run on edge devices.
His interdisciplinary approach, based on optimizing neural networks, contributes to scalable technologies
addressing social needs.
Date: Monday, August 4, 2025
Time: 1:00 – 3:00 pm
Location: CAMP 194
Join Link: https://clarkson.zoom.us/j/92393963947?pwd=h4hUTX6t1eDhrHMpSIgcXfXauB0uVq.1
Committee members:
Dr. Masudul H. Imtiaz (Chair)
Dr. Daqing Hou
Dr. Abul Khondker
Dr. Chen Liu
Dr. Kevin Fite