SmartJump: A Continuous Jump Detection Framework on Smartphones

Performing jump exercise can maintain a healthy lymphatic system, which keeps human body in an optimal condition and is a critical component of human immune system. Accurate jump detection and count are crucial to patients with a dysfunctional lymph system. In this article, we present a continuous jump detection framework on smartphones, SmartJump, for human jump detection and count, by leveraging the accelerometer and magnetometer ubiquitously built into smartphones. Specifically, SmartJump collects sensing data from the accelerometer and magnetometer, and processes these data through coordinate system translation and data smoothing filter. Then, jump features are extracted based on the smoothed z-axis acceleration data using the peak and valley detection algorithm and then are matched with the concluded three features from the analysis of physical jumps using a finite state machine for jump detection and count. We implement SmartJump on Samsung S6 Edge smartphones and recruit six subjects for data collection. We evaluate the accuracy of SmartJump in terms of five-fold cross-validation test, self-test, and leave-one-out cross-validation test, and the experimental results indicate that SmartJump achieves an average of 96.4 % recall, 97.2 % precision, and 96.8 % F1 score in five different scenarios.