Accelerometers record acceleration in three orthogonal directions relative to the Earth gravity. The left figure shows the 80Hz data collected from 2 subjects'walking. The x-axis is recording time in seconds and the y-axis is the acceleration vector in g units. We introduced novel statistical methods to predict the activity types of one person at sub-second resolution using the labeled activity data from others. A key challenge in the across subject movement prediction is that trixial accelerometry data might not be comparable because accelerometers might be rotated or tilted and also have some small bias; see the top panels for an illustration. We introduced a simple procedure to normalize the data and now the data look more similar across subjects; see the bottom panels. For more details on the normalization and how the prediction was done, please refer to my tentatively-accepted-by-Biometrics paper Movement prediction using accelerometers in a human population.