Running, walking, sitting or biking? - Motion prediction with acceleration and rotation
summary
Smartphone and USB IMU logs at 50 Hz produce acceleration (3 axes) and rotation rate (3 axes) time series and the pipeline extracts absolute acceleration (sqrt sum squares) and a rolling max−min window plus an FFT to pick an FFT amplitude peak from one‑second windows. Supervised labels from file names train an SVC with linear kernel on an 80/20 split, giving near‑perfect score on recorded data; a Tinkerforge hardware demo shows real‑time feature calc and live prediction with SciKit‑Learn and Pandas.
AI-written summary · about
chapters (10)
- 0:00Introduction
- 1:19Activity prediction features
- 4:12Absolute acceleration
- 6:22Acceleration variance
- 7:05Rotation rate signals
- 8:29FFT for frequency
- 9:19Model training accuracy
- 11:48Real-world Tinkerforge test
- 13:37Live classification demo
- 16:05Q&A
related talks
34mMining smartphone sensor data with python
06mApproaching the Real-World: Supporting Activity Recognition Training with Virtual IMU Data
33mAnalytics in the age of the Internet of Things #bbuzz
06mIMUTube: Automatic extraction of virtual on-body accelerometry from video for human activity ...
52mMuvr
27mDissecting ActivityRecognition with Multi-Sensor Simulation and Its Practical Deployment
06mRobust Inertial Motion Tracking through Deep Sensor Fusion across Smart Earbuds and Smartphone
32mHacking Fitness