imt-ikarus 1.2.1

IKArus - Inertial and Optical Data of Kinematic ChAin Motion

IKArus Inertial and Optical Data of Kinematic ChAin Motion

This repository hosts the Ikarus dataset. It contains real-world IMU and OMC data from kinematic chain motion.

It contains:

• dataset/*: the preprocessed dataset
• preprocess/*: unpreprocessed data and preprocessing logic to re-create the dataset folder
• src/*: a Python Package that
  • provides convenient access to the data from a Python environment
  • defines benchmarks
  • defines baselines

dataset/*

There exist two different five-segment kinematic chains, namely arm and gait. The two five-segment kinematic chains differ in types of joints in the setup.

• arm: S(egment)1 - 1D joint (Z) - S2 - 1D joint (Y) - S3 - 1D joint (X) - S4 - 3D joint - S5
• gait: S2 - 1D joint (Y) - S3 - 2D joint - S4 - 2D joint - S5 - 1D joint (Y) - S1

For each of the two kinematic chains several experiments expX (recordings) have been performed. An experiment is one continous recording of IMU and OMC data whilst the kinematic chain is moving.

Each experiment contains several different types of motions motionY, such as, e.g., fast motion, slow motion, standstills...

The motions are from one continous experiment and are ordered chronologically. As a result, two (or more) motions, e.g., motion01 and motion02 can be concatenated to create one longer continous timeseries.

The dataset folder hierarchy is structured in this way:

dataset
├── arm
│   ├── exp01
│   │   ├── motion01_canonical
│   │   │   ├── exp01_motion01_imu_nonrigid.csv
│   │   │   ├── exp01_motion01_imu_rigid.csv
│   │   │   └── exp01_motion01_omc.csv
│   │   ├── motion02_pause1
│   │   ├── motion03_slow1
│   │   ├── motion04_pause2
│   │   ├── motion05_fast
│   │   ├── motion06_pause3
│   │   ├── motion07_fast_slow_fast
│   │   ├── motion08_freeze1
│   │   ├── motion09_fast_slow
│   │   ├── motion10_freeze2
│   │   ├── motion11_slow2
│   │   ├── motion12_shaking
│   │   └── motion13_pause4
│   ├── exp02
│   ├── exp03
│   ├── exp04
│   └── exp05
├── gait
│   ├── exp06
│   ├── exp07
│   ├── exp08
│   ├── exp09
│   ├── exp10
│   └── exp11

For each experiment expX and motionY the folder armOrGait/expX/motionY contains three files:

• expX_motionY_imu_nonrigid.csv: The acc/gyr/mag measurements of five IMUs where each IMU is nonrigidly attached to the respective segment.
• expX_motionY_imu_rigid.csv: The acc/gyr/mag measurements of five IMUs where each IMU is rigidly attached to the respective segment.
• expX_motionY_omc.csv: For each of the five segments, four 3D marker positions and the absolute orientation as a quaternion.

All .csv files contain the respective sampling frequency (e.g. 40 Hz) as the first line:

# sampling frequency: 40

The header of, e.g. expX_motionY_imu_rigid.csv, starts with:

seg1_acc_x,seg1_acc_y,seg1_acc_z,seg1_gyr_x,seg1_gyr_y,seg1_gyr_z,seg1_mag_x,seg1_mag_y,seg1_mag_z, ...

[!IMPORTANT]
The identifiers of the segments for arm are: seg1 - seg2 - seg3 - seg4 - seg5, and for gait are: seg2 - seg3 - seg4 - seg5 - seg1

The identifiers of the markers (M1 to M4) for the five segments (S1 to S5) can be read of from the following figure

preprocess/*

Refer to the readme in the subfolder.

src/*

Requires Python 3.10 or higher.

The Python Package allows convenient access to the data and can be installed with

pip install imt-ikarus

Note that the Python Package will automatically download the required data on-demand. By default it comes with no data at all and there is no data stored in a redundant way.

Quickstart

import ikarus

# concatenates 3 motions: the motion01, motion02 and motion03 
data = ikarus.load_data(
    exp_id         = 1,
    motion_start   = 1,
    motion_stop    = 3,
    resample_to_hz = 100
)

print(data.keys())
# ['seg1', 'seg2', 'seg3', 'seg4', 'seg5']

print(data['seg1'].keys())
# ['imu_rigid', 'imu_nonrigid', 'marker1', 'marker2', 'marker3', 'marker4', 'quat']

print(data['seg1']['imu_rigid'].keys())
# ['acc', 'gyr', 'mag']