# Egait Adidas 2014 - A validation dataset for foot-worn IMU-based gait analysis systems
This dataset contains IMU data from 20 participants with motion capture reference.
The data was acquired in the laboratory of the adidas AG in Herzogenaurach, Germany
The two IMU sensors were mounted laterally on the shoes.
Depending on the trail, these sensors were either Shimmer3 sensors (sampling-rate of 204.8 Hz) or Shimmer2R sensors (sampling-rate of 102.4 Hz).
Each participant performed multiple trials with different stride length and gait speeds and the different sensors.
The reference system was a Vicon motion capture system (sampling-rate 200 Hz) with 16 cameras tracking 6 markers at each show.
Both systems were synchronized using a trigger signal generated by a light-barrier at the beginning of the capture volume.
For more details see the following two publication:
Kanzler, Christoph M., Jens Barth, Alexander Rampp, Heiko Schlarb, Franz Rott, Jochen Klucken, and Bjoern M. Eskofier. “Inertial Sensor Based and Shoe Size Independent Gait Analysis Including Heel and Toe Clearance Estimation.” In Proceedings of the Annual International Conference of the IEEE Engineering in Medicine and Biology Society, EMBS, 2015-Novem:5424–27, 2015. https://doi.org/10.1109/EMBC.2015.7319618.
Hannink, Julius, Malte Ollenschläger, Felix Kluge, Nils Roth, Jochen Klucken, and Bjoern M. Eskofier. “Benchmarking Foot Trajectory Estimation Methods for Mobile Gait Analysis.” Sensors 17, no. 9 (August 23, 2017): 1940. https://doi.org/10.3390/s17091940.
## Folder structure
This data release contains two folders:
- `data`: contains the IMU data as a binary format, the extracted gait parameters (from the original publication) as plain text format and the corresponding Vicon data as c3d files.
- `calibrations`: contains the calibration data for the IMU sensors, that can be used to calibrate the raw IMU data
NOTE: While we included the calculated gait parameters from the original publication for completeness, we recommend only using the raw IMU data and calculate the gait parameters yourself, as the calculation of the original gait parameters can not be fully reproduced and was performed with different calibration files than the ones provided here.
## Data format
The used data formats are custom binary formats and can be loaded using the `gaitmap_datasets` python library.
This library also handles the calibration of the IMU sensors and correctly loading all event data.
The library can be installed via pip:
```bash
pip install gaitmap_datasets
```
And an example on how to specifically load the Egait-Adidas dataset can be found in the [here](https://mad-lab-fau.github.io/gaitmap-datasets/auto_examples/egait_adidas_2014.html#sphx-glr-auto-examples-egait-adidas-2014-py).
To use the library with the dataset, download and unpack the entire dataset in a folder of your choice.
Then you can load the data by specifying the path to the folder when loading the dataset (see below), or by creating a dataset config file as explained in the README of the `gaitmap_datasets` library.
```python
from gaitmap_datasets import EgaitAdidas2014
dataset = EgaitAdidas2014(path_to_dataset_folder)
```
## Coordinate systems
See the original paper and our [usage example](https://mad-lab-fau.github.io/gaitmap-datasets/auto_examples/egait_adidas_2014.html#sphx-glr-auto-examples-egait-adidas-2014-py) in the `gaitmap_datasets` library for more details on the coordinate systems.
