Evaluating the validity of walker sensors for posture monitoring

System architecture

System description

The data adquisition system is comprised by three main subsystems:

• Walk-it rollator: a low cost smart rollator developed by authors.

• MoCap system: a 10 Optitrack Prime x13 camera setup connected to ROS using MOCAP4ROS2 framework and configured to track both the rollator and user using a 13 marker configuration based on COCO18.

• Empatica E4 wristband: a bluetooth-enabled wearable with four sensors: an electrode for Electrodermal activity (EDA), a 3-axis accelerometer, a temperature sensor, and a photoplethysmography (PPG) to measure blood volume pulse (BVP) from which it derives HR and the inter beat interval (IBI).

All four subsystems publish data on real time using ROS2, and rely on its recording capabilities to obtain a sinchronized dataset. ROS2 transport is done using an ethernet network, with a WiFi extender to connect the rollator. Empatica E4 requires the use of an android device as gateway to reach the network, as EmpaLink Software is not available in other platforms.

Sensors layout

In order to perform the tests, users have to equip with an E4 wristband and 13 reflective markers for the MoCap system. For the MoCap system, COCOA markers related with head posture have been removed, focusing on those relevant for motion. Fig. 1 depicts a volunteer with all sensors attached.

Experiment description

Experiments were performed at IMECH Institute in Málaga. Volunteer was asked to walk using the rollator along different trajectories. These trajectories were also performed both freely and mimmicking some pathological deviations. 

Dataset structure

Original ROS2 bags have been processed in order to make datasets independant from the robotic framework. Each experiment dataset has been split into 19 comma separated (csv) files and compressed as an independant zip file. These are the filenames inside each zip, related to one specific parameter:

Future work

This comprehensive dataset offers simultaneous information on MoCap data, physiological metrics, gait and support data. It provides a unique opportunity for in-depth analysis of the interactions between physiological responses and gait mechanics from multiple sensors. By integrating these diverse data sources, potential correlations between the different sensors can be explored and analyze to what extent it is possible to have an estimate of most parameters using just those sensors embedded on a walker. Future work will focus on extending this dataset to more users and configurations and leveraging this dataset to validate the use of simplified sensor systems for real-world applications.

Acknowledgements

 The authors would like to thank to IBIMA and IMECH Institutes for letting us to use their MoCap systems.