top of page

Project Outcomes

Ground Truth Project

Screen Shot 2020-04-27 at 12.46.34 PM.pn

  • Selected model: Exponential Gaussian Process Regression   

    • Hold out Validation 20%

    • Kernel Scale : 2135

    • Basis Function : Constant 

    • Sigma : 0.0291

    • Beta : 0.7608

    • Holdout Validation : 20%

    • RMSE : 0.85493​

  • AI for stance time is a viable option for walking gait analysis

  • The trained regression model can predict gait asymmetry using segmented acceleration signals

​

Acceleration Project 

Screen Shot 2020-04-27 at 12.46.19 PM.pn

  • ​

    • The segmentation of gait cycles in acceleration data is possible without the GRF

    • Accelerometer data (alone) has the ability to successfully predict gait asymmetry which could be integrated into a wearable sensor for continuous detection

    • Future improvements will allow users to predict asymmetry using a wearable accelerometer sensor

bottom of page