The scenario
A team building a wearable fall detection model needs to expand a small labeled sensor dataset. This flowchart shows how raw accelerometer windows are labeled and augmented by rotation and scaling before model training.
What is in this drawing
Read the decisions behind it.
01
Using 3D rotation to simulate device orientation changes
02
Using magnitude scaling to simulate varying intensity
03
Merging fall and non-fall augmented copies into one balanced dataset
Reusable when designing synthetic data pipelines for inertial sensor classification, especially when collecting real fall events is expensive or risky.
More like this