Title: Towards novel sensing methodologies for the ubiquitous assessment of joint kinematics: 

In the field of medical sensing, how can tuning the intended accuracy based on application and anatomical constraints, reduce the intrusiveness of the required sensing hardware?

Fereshteh Shahmiri

Ph.D. Student, Computer Science

School of Interactive Computing

Georgia Institute of Technology

Date: Friday, October 29, 2021

Time: 3:00 PM - 5:00 PM (EST)

Location (remote via BlueJeans): https://bluejeans.com/269896328/8953

Committee:

Dr. Keith Edwards (Advisor), School of Interactive Computing, Georgia Institute of Technology

Dr. Omer Inan (Co-Advisor), School of Electrical and Computer Engineering, Georgia Institute of Technology

Dr. Elizabeth Mynatt, School of Interactive Computing, Georgia Institute of Technology

Dr. Thad Starner, School of Interactive Computing, Georgia Institute of Technology

Dr. Gierad Laput, Apple Inc. 

Abstract:

Our body movements and postures are key indicators of our health and well-being. In many cases in our daily lives, those movements and postures need accurate, low-cost, long-term, and unobtrusive monitoring and quantitative assessments. Hence, many motion tracking technologies including wearable and portable sensing devices are developed for a variety of medical applications. The nature and complexity of the developed sensing systems vary case by case, with respect to the requirements that every application impose on the system. For instance, one application may track a single joint with a large range of motions (RoM), like knee joint, that is accompanied by sophisticated deformations of soft tissues and skeletal muscles around the joint, while another application may necessitate the characterization of complex motions from a group of small-size joints like finger joints. Regardless, the joint kinematics and soft tissue artifact (STA) that are highly variable relative to different body segments, subjects, and types of motor activities, cause challenges in the development of accurate and usable systems. Given these challenges in the field of medical sensing, we adopt a unique approach to bridge the gap in the literature. We propose a set of novel sensing methodologies that allow tuning the recognition/ tracking accuracy based on application and anatomical constraints to reduce the intrusiveness of the required sensing hardware. We invent four novel sensing platforms and employ two implementation approaches to assess joint kinematics by either directly estimating the position and orientation (pose) of a joint or classifying joint(s) motions through mapping the motion-related features to an activity set.  We demonstrate and evaluate our systems through multiple applications including the assessment of finger(s) dexterity, jaw motions and oral activities , and limited or abnormal joint mobilities (e.g., knee joint, or spine vertebrae).