An Integration-Free Accelerometer Based Approach for Estimating Peak Knee Angle

Overview

Knee angle tracking has important applications in rehabilitation and sports training. In this project, a wearable sensor system is developed for using inertial measurement units (IMUs) to track knee angles. It consists of two IMU sensors, an Arduino Nano board, and a Micro SD module. In addition, a new integration-free accelerometer- based method for tracking peak knee angle is proposed. The method takes the advantage of the fact that gravity is the only acceleration at the peak of stride. The peak knee angle is the angle formed by the thigh and shank when the knee reaches its peak position in a walk stride. It is an important parameter for measuring the range of motion. Experiments were conducted with three different users to evaluate the accuracy of the developed sensor system and the proposed peak knee angle tracking method. Two additional tracking methods, an optical tracking method and a conventional gyroscope-based method, were also used in the experiments for comparison purposes. Experimental results show that the developed sensor system and the proposed method can accurately track peak knee angles. The average root-mean-square (RMS) error of the three users’ peak knee angles obtained by the proposed method is 7.6 . It also shows that the proposed method is more robust than the conventional gyroscope-based method. Minimizing Error in Machine Learning Algorithms by Adjusting Model Hyperparameters Charlotte White Lakenheath High School Machine learning (ML) allows a computer system to detect patterns in a set of data and make predictions. ML techniques can be applied to many fields both inside and outside of computer science, such as economics or healthcare. The researcher’s focus in this study is not on the prediction of specific data but rather on the optimization of the machine learning model itself in order to produce more accurate predictions based on any data set. The researcher analyzed the resulting error of twenty-seven combinations of the values of three hyperparameters or setup parameters (batch size, step size, and learning rate) when applied towards the prediction of median house values given the population in a compact geographical area using data from the 1990 U.S. Census. While the researcher hypothesized that each individual variable would have a direct, independent effect on the resulting error, the data demonstrated that this is partially true, but also that two of the input variables are interdependent in producing ideal results. This study produced nine combinations of the three variables that yielded optimal predictions, while simultaneously demonstrating the complexity of the optimization of machine learning models.

Competition history

• JSHS 2020 Category not listed

Resources

• National JSHS abstract book ↗

Related projects

Closest projects by meaning, across every fair and year in the corpus.