Smart Medical Devices Lab (SMDL)

February 6, 2023

Yoon speaks at Husky Highlights seminar

The Offices of Sponsored Research and Connected Learning would like to invite you to the Husky Highlights Seminar on Tuesday featuring two great presentations.

Tuesday, February 7, 2023, from 3:30-5:00PM

Towards high-quality inference services in mobile edge computing

 –Dr. Yang Peng

Deep neural networks (DNN) have enabled dramatic advancements in mobile applications such as real-time video analytics, speech recognition, and autonomous navigation. In order to facilitate mobile devices in executing complex inference jobs under strict latency requirements, edge intelligence has been proposed to offload DNN inference tasks from end devices to more powerful edge servers for accelerated and high-quality inference. However, state-of-the-art edge intelligence solutions still need to overcome the challenges of interrupted and downgraded inference quality, given continual user and edge server movement. In this project, we designed and developed various innovative schemes to help improve the performance of edge intelligence, leveraging techniques that include model selection, dynamic batching, frame resizing, and server handover handling. Our solutions have been implemented and evaluated via extensive simulation and testbed experiments, demonstrating promising advancement over existing solutions. These achieved results also laid the ground for our ongoing research in vehicular edge computing.

Gizmo 2.0 – a cross-disciplinary robot for enhancing recovery after stroke-induced hemispatial neglect

 –Dr. W. Jong Yoon

Stroke induces a variety of functional impairments, including motor function, depression, and hemispatial neglect (HSN). Our team’s long-term central question is, does there exist a way of accelerating its recovery, beyond that provided by current therapy?  To solve the main problem, three parties synergistically worked: design and development of robots (Gizmo 1 and Gizmo 2), cloud interface research, and sensory feedback collection/processing (EEG and camera). Thanks to a grant-funded workforce, the team has been able to finalize a pilot project (Gizmo 1) for upper arm rehabilitation and we are waiting to soon perform a final human test. We are in the middle of assembling our final product generation of Gizmo 2.0, designed for HSN patients, with sufficient functionality to demonstrate its ability to simultaneously (a) interface with EEG and (b) track face, (c) provide visual, auditory and tactile feedback, (d) as tested on healthy volunteers in 2023. With participants from computer science, engineering, industrial design, and health studies, we offered ourselves as a model of cross-disciplinary scholarship.