Note
Click here to download the full example code
EgaitSegmentationValidation2014 - A Stride Segmentation validation dataset#
The EgaitSegmentationValidation2014 dataset allows access to the stride segmentation validation dataset recorded for the EGait system. It contains multiple 4x10 m walks and simulated “free-living” walks recorded by two foot worn IMU sensors.
Two sets of reference stride borders are provided for the dataset:
1. The orignal stride borders from the original publication. These stride borders were labeled manually by multiple gait experts looking at the raw IMU signal and a video of the participant. The gait experts specifically only labeled full straight strides 2. Updated stride borders, which contain the original stride borders and additionally contain stride borders for all turn and stair strides in the dataset. These new annotations where only performed on the raw data.
General information#
The dataset was recorded with Shimmer 2R sensors. In these IMU nodes, the coordinate systems of the accelerometer and the gyroscope are different.
In the version provided in this dataset, we fix this by transforming the gyroscope data to the accelerometer coordinate system and then transform the combined data to the coordinate system of the gaitmap coordinate system.
Warning
The calibration files distributed with the dataset are likely of low quality. We recommend to only use this dataset for validation of stride segmentation algorithms. Algorithms for spatial parameters that depend on the exact values of the IMU, might not provide good results with this dataset.
Warning
For this example to work, you need to have a global config set containing the path to the dataset.
Check the README.md for more information.
from gaitmap_datasets import EgaitSegmentationValidation2014
First we will create a simple instance of the dataset class.
Based on the index you can select either a specific cohort or test, or a specific participant.
only_free_walk = dataset.get_subset(test="free_walk")
only_free_walk
We will investigate the data for a single participant in the following for both types of tests
free_walk = only_free_walk.get_subset(participant="GA214030")
free_walk
Free-walk#
During the free-walk tests participants were asked to perform a series of activities. This mostly consisted of walking around a room and up and down stairs. We will plot the data together with the stride labels. We can see that multiple strides were labeled over the 4-min period of the measurement. The only exception is a small signal region in the center
import matplotlib.pyplot as plt
def plot_strides(imu_data, segmented_stride_list):
fig, axs = plt.subplots(2, 1, sharex=True)
foot = "right_sensor"
imu_data[foot].filter(like="acc").plot(ax=axs[0])
imu_data[foot].filter(like="gyr").plot(ax=axs[1])
for i, s in segmented_stride_list[foot].iterrows():
s /= free_walk.sampling_rate_hz
for ax in axs:
ax.axvline(s["start"], color="k", linestyle="--")
ax.axvline(s["end"], color="k", linestyle="--")
return fig
imu_data = free_walk.data
segmented_stride_list_original = free_walk.segmented_stride_list_original_
segmented_stride_list = free_walk.segmented_stride_list_
If we plot the original stride list (the one without stair strides labeled), we can see that there is a section in the middle without any labels
fig = plot_strides(imu_data, segmented_stride_list_original)
fig.title = "Original stride list"
fig.show()
If we zoom into this region, we can see that the signal looks “gait-like”.
However, this corresponds to stair walking, which was explicitly not labeled as a stride by the authors of the
dataset, as they wanted to show that the algorithms they developed could differentiate between stair walking and
level walking.
Further, all turning strides are not labeled when using the original stride list (
segmented_stride_list_original_.
fig = plot_strides(imu_data, segmented_stride_list_original)
fig.axes[0].set_xlim(145, 165)
fig.show()
The new relabeled stride list (segmented_stride_list_) contains all strides, including the stair and turning
strides.
If we plot this stride list, we can see that the signal region in the middle is now labeled as strides.
fig = plot_strides(imu_data, segmented_stride_list)
fig.title = "New stride list"
fig.show()
4x10 m walk#
only_gait_test = dataset.get_subset(test="4x10m")
gait_test = only_gait_test.get_subset(participant="GA112030E3")
gait_test
We will plot the data together with the manually labeled strides. We can clearly see the 4 straight walks during the test.
Like before, if we use the original stride list, the turning strides in between the bouts are not labeled.
imu_data = gait_test.data
segmented_stride_list_original = gait_test.segmented_stride_list_original_
segmented_stride_list = gait_test.segmented_stride_list_
fig = plot_strides(imu_data, segmented_stride_list_original)
fig.title = "Original stride list"
fig.show()
If we use the new stride list, we can see that the turning strides are now labeled.
fig = plot_strides(imu_data, segmented_stride_list)
fig.title = "New stride list"
fig.show()
Stride List Recommendation#
While in many cases we are only interested in analyzing straight strides, we recommend to use the new stride list when validation stride segmentation algorithms. It is usually better to have a segmentation algorithm with high sensitivity that is able to identify all stride-like signal portions and then filter out unwanted stride types in later processing steps. For this reason, we made the new stride list the default stride list in the dataset.
Total running time of the script: ( 0 minutes 7.407 seconds)
Estimated memory usage: 85 MB