Accuracy of Posture Estimation by ActiGraph and Development of Posture Prediction Model from Raw Acceleration Data

Full body motion tracking system was used to obtain accurate posture information under free-living condition and the accuracy of posture prediction by ActiGraph was verified. ActiGraph tends to detect people as standing when they are actually sitting. By combining a prediction model that detects posture change, lying posture, and walking from raw acceleration data obtained by ActiGraph, we were able to predict posture information with higher accuracy.Clinical Relevance- By combining model for predicting posture changes, lying posture, and walking from raw acceleration data obtained by ActiGraph, it is possible to obtain more accurate posture information.