Learning Subtle Mo0on Cues For Human Performance Analysis With Wearable Sensors

Abstract

In this report, we focus on three projects. The first one involves a weakly supervised machine learning framework, called multiple instance learning (MIL), for automatic detection of Parkinson’s Disease motor symptoms in daily living environments. Our primary goal was to develop a monitoring system capable of being used outside of controlled laboratory settings. The goal was to track the medication cycles at home and provide valuable clinical feedback. Most of the relevant prior works involve supervised learning frameworks (e.g., Support Vector Machines), which are inadequate for handling incomplete ground truth information that are inherent to daily living environments. We monitored two Parkinson’s disease (PD) patients, each for four days with a set of five tri-axial accelerometers and utilized a MIL algorithm based on axis parallel rectangle (APR) fitting in the feature space. We were able to detect subject specific symptoms (e.g. dyskinesia) that conformed to a daily log maintained by the patients. In the second project, we have developed a quantitative framework for skill assessment of paramedics performing endotracheal intubation (ETI), which is a critical emergency medicine procedure. In our pilot study, we developed a quantitative framework for discriminating the kinematic characteristics of providers with different experience levels. The system utilizes statistical analysis on spatio-temporal multimodal features extracted from optical motion capture, accelerometers and electromyography (EMG) sensors. Our experiments involved three individuals performing intubations on a dummy, each with different levels of training. Quantitative performance analysis on multimodal features revealed distinctive differences among different skill levels. In the third project, we attempt to build a system that will provide a quantitative measure of the quality of the performance of a therapeutic home exercise. This system performs assessment of human motion quality using body-worn tri-axial accelerometers. This system will have important clinical benefit because patients diagnosed with knee osteoarthritis are prescribed therapeutic exercises to be completed in the home and they often fail to adhere to the prescribed program or perform the exercises incorrectly. The results show that we could use accelerometer data and simple multi-label classification techniques to recognize subtle errors in exercise performance.

Award ID: 0931999

  • 0931999
  • CPS Domains
  • Medical Devices
  • Quantitative Verification
  • Systems Engineering
  • Real-Time Coordination
  • Health Care
  • Validation and Verification
  • CPS Technologies
  • Education
  • Foundations
  • National CPS PI Meeting 2012
  • 2012
  • Poster
  • Academia
  • CPS PI MTG 12 Posters & Abstracts
Submitted by Jessica Hodgins on