EgaitParameterValidation2013 - Spatial Parameters#

Further Links:

A challenge for stride-by-stride comparison of spatial parameters on straight walks.

General Information#

Dataset

The EgaitParameterValidation2013 dataset [1] (usage example, download) contains short walks of 100 geriatric patients with a GaitRite carpet as a reference system.

Sensor System

Two IMU sensors (Shimmer 2, 102.4 Hz) are attached laterally to the shoes of the participants.

Reference System

A GaitRite carpet (100 Hz) is used as a reference system. The reference system was manually synchronized with the IMU sensors on a stride level, by counting the strides in both systems. Further, the dataset contains manually annotated stride borders following the definition of Barth et al. [2].

Implementation Recommendations#

The IMU data is cut to the region on the carper, where the reference system is valid. This means, that the data starts mid-movement and there is no resting period within the signal.

The start and end value of each stride in the provided segmented stride list follow the stride definitions by Barth et al. [2]. This means the start and end-values are defined by the minimum in the gyr_ml axis right before the toe-off. The ground truth stride length is calculated from the marker position at the heel marker by calculating the distance traveled by this marker in the ground plane between the heel strike right before a stride and the heel strike within the stride. This means for the first stride in the segmented stride list and for each stride after a break, no ground truth stride length can be calculated. The pipeline should also not calculate parameters for these strides. To correctly handle these shifts in stride definition, you can use convert_segmented_stride_list or check the stride ids in the calculated ground truth parameters to remove strides without ground truth from your calculations. Further check the dataset example for more guidance on this. The final calculated parameters should match the provided stride list and should have the same stride ids. Note, that we assume that parameters are calculated for each stride. Missing strides are not handled by the evaluation.

References#

[1]

Rampp, Alexander, Jens Barth, Samuel Schuelein, Karl-Gunter Gassmann, Jochen Klucken, and Bjorn M. Eskofier. ] “Inertial Sensor-Based Stride Parameter Calculation From Gait Sequences in Geriatric Patients.” IEEE Transactions on Biomedical Engineering 62, no. 4 (April 2015): 1089-97. https://doi.org/10.1109/TBME.2014.2368211.

[2] (1,2)

Barth, Jens, Cäcilia Oberndorfer, Cristian Pasluosta, Samuel Schülein, Heiko Gassner, Samuel Reinfelder, Patrick Kugler, et al. “Stride Segmentation during Free Walk Movements Using Multi-Dimensional Subsequence Dynamic Time Warping on Inertial Sensor Data.” Sensors (Switzerland) 15, no. 3 (March 17, 2015): 6419-40. https://doi.org/10.3390/s150306419.