Efficient Computational Methods for Understanding Back Move-ment and Pain from Dynamic Data Modeling
This project uses machine learning algorithms to better understand back movement and low back pain. We apply supervised learning time series algorithms to data collected from Backtracks’ wearable de-vice — which consists of a malleable think curve that reads data collected from the participants’ spine movements. At each time step, such movements are represented as a curve; the dynamic evolution of this curve in time represents an individual’s spinal movements. The goal of this project is to use time series analysis to detect the type of activity (e.g., walking, sitting, bending forward) the participant is doing at any particular window of time. We will use time delay embeddings (Frank, Mannor et al. 2010) on individual observations from the Backtrack device in order to represent the dynamics and support vector ma-chines for the classification tasks.