A multi-sensor gait dataset collected under non-standardized dual-task conditions

Liao, Yuanyuan1 ; Cao, Junjie1 ; Yu, Lisha2 ; Xiang, Jianbang1 ; Zhao, Yang 1

Research facility: MOCAP,Kinect,IMU

Published Apr 25, 2025 on Dryad. https://doi.org/10.5061/dryad.2rbnzs7z3

Data files

Apr 25, 2025 version files 267.75 MB

Data_new.zip

267.74 MB
README.md

8.47 KB

Abstract

Non-standardized dual-tasks have recently gained attention in health monitoring and post-operative rehabilitation. By collecting data with multiple sensors, we can quantify motion characteristics from different perspectives and explore the complementarity and interchangeability between sensors. Currently, there is a lack of publicly available non-standardized dual-task gait datasets collected with multiple sensors, thus we proposed a dataset (NONSD-Gait) from 23 healthy adults walking back and forth over 7 meters under three dual-task conditions, collected by three types of sensors: optical motion capture (MOCAP) system, depth camera and inertial measurement unit (IMU). MoCap captured the 3D trajectories of 22 markers attached to the subject using 8 optical cameras, while the depth camera recorded the 3D trajectories of 25 joints through a non-contact depth camera. The IMU was placed on the left ankle to record 3-axis acceleration and angular velocity data. Each participant underwent two repeated experiments for each task. Moreover, this dataset also includes extracted spatio-temporal gait parameters and kinematic parameters, supporting gait feature recognition in complex scenarios and multimodal gait data analysis.

https://doi.org/10.5061/dryad.2rbnzs7z3

Description of the data and file structure

File: Data.zip

Description:

The "Raw" folder contains the complete experimental data collected by the three sensors.
The "Processed" folder includes segmented data from the three sensors.
The "Parameters" folder contains the spatio-temporal gait parameters and kinematic parameters extracted separately by the three sensors.
The demographic information of the subjects is stored in a file named **Demographics.csv. **All participants signed informed consent and agreed to the open publication of the data. This study does not involve sensitive data.

Each participant’s folder is named: sampleID, with ID ranging from 01 to 23.

The folders for the two repeated experiments are named with timeID, with ID ranging from 1 to 2.

The tasks are named task_name, with names being normal, cup, text, and web.

Task_normal: Walking at a normal speed.
Task_cup: Walking while holding a cup filled with water.
Task_text: Walking while texting.
Task_web: Walking while browsing a webpage.

The folders for each phase of straight walking and turning after data segmentation are named go, back, and turn.

Raw data:

The four folders are MOCAP, Kinect, IMU and Timestamp. Each sample folder contains folders for two repeated experiments. Each experiment folder contains four task files named task_normal.csv, task_cup.csv, task_text.csv and task_web.csv. The Timestamp folder contains 23 sample folders, each of which includes timestamp files for two repeated experiments named timeID.csv. The table structures for the data from each sensor and the manual timestamps are as follows:

1) MOCAP: Data matrix with scale n × 69. n is the number of frames. The first column is frame number, the second column is relative time, the third column is the timestamp, and the following 66 columns are the three-dimensional position variables of 22 reflective markers.

Unit description of MOCAP: All coordinate values in the X, Y, and Z directions are expressed in millimeters (mm).

Notes: Since we started recording on the device in advance, the raw data contains a large number of blank values at the beginning. In the middle part, brief missing values in the marker coordinates may occur due to self-occlusion. These issues have already been addressed in the "Processed" dataset.

2) IMU: Data matrix with scale n × 15. n is the number of frames. The first column is the timestamp, the second column is the name of IMU, and the next 13 columns are variables such as three-axis acceleration, three-axis angular velocity, three-axis angle, three-axis magnetic field and temperature.

Unit description of IMU: The unit of three-axis acceleration is g (where 1 g ≈ 9.81 m/s²), three-axis angular velocity is in degrees per second (°/s), three-axis angle is in degrees (°), three-axis magnetic field is in microtesla (μT), and temperature is in degrees Celsius (℃).

3) Kinect: Data matrix with scale n × 76. n is the number of frames. The first column is the timestamp, and the following 75 columns are three-dimensional position variables for 25 joints.

Unit description of Kinect: All coordinate values in the X, Y, and Z directions are expressed in meters (m).

Notes: Similarly, due to the early start and delayed stop of the recording device, there are a large number of missing values at the beginning and end of the data. These missing values are denoted as "None-tracked". These issues have already been addressed in the "Processed" dataset.

4) Timestamp: Data matrix with scale 4 × 7. The 4 rows represent the tasks. The first column is the task name, and the remaining 6 columns are the recorded 6 time points.

Processed data:

The processed data folder includes five sensor folders: MOCAP_5m, MOCAP_2m, IMU_5m, IMU_2m, and Kinect_2m. Each sensor folder contains two experiment folders. Each experiment folder contains three process folders. Each process folder contains four task folders. Each task folder contains 23 sample files named sampleID_task_name.xlsx.

Unit description of MOCAP and Kinect: All coordinate values in the X, Y, and Z directions are expressed in meters (m).

Parameter extraction data:

This folder includes five sensor folders: MOCAP_5m, MOCAP_2m, IMU_5m, IMU_2m, and Kinect_2m. Each folder of MOCAP and Kinect contains two files named spatio-temporal_parameter.csv and kinematic_parameters.csv. Each folder of IMU contains one file, named kinematic_parameters.csv. The structures of the extracted parameter tables are as follows:

1) Spatio-temporal gait parameters for MOCAP and Kinect: Data matrix with scale n × 14. n is the number of walking segments. The first column is the sensor name, the second column is the sample ID, the third column is the experiment session, the fourth column is the walking phase, the fifth column is the task name, and the remaining nine columns are the means of the extracted spatio-temporal gait parameters.

Unit description: All length units are in meters (m), time is in seconds (s), and velocity is in meters per second (m/s).

2) Kinematic parameters for MOCAP and Kinect: Data matrix with scale n × 65. n is the number of walking segments. The first column is the sensor name, the second column is the sample ID, the third column is the experiment session, the fourth column is the walking phase, the fifth column is the task name, and the remaining 60 columns are the statistical measures of the extracted kinematic parameters.

Unit description: All velocity units are in meters per second (m/s), and all angle units are in degrees (°).

3) Kinematic parameters for IMU: Data matrix with scale n × 113. n is the number of walking segments. The first column is the sensor name, the second column is the sample ID, the third column is the experiment session, the fourth column is the walking phase, the fifth column is the task name, and the remaining 108 columns are the statistical measures of the extracted kinematic parameters.

Unit description: The unit of three-axis acceleration is g (where 1 g ≈ 9.81 m/s²), three-axis angular velocity is in degrees per second (°/s) and three-axis angle is in degrees (°).

Code/software

Clone repository

https://github.com/Vanessa-lyy/Multi-source-gait-dataset.git
Create environment with anaconda3
Download dataset
Modify the file path in the code

Human subjects data

All human subjects included in this dataset provided explicit written informed consent for their de-identified data to be published in a public repository. The data have been de-identified to protect participants' privacy.

The following steps were taken to ensure de-identification:

All direct identifiers such as names, identification numbers, contact information, and other personally identifiable information (PII) have been completely removed;
No dates of birth, geographic information, or unique codes that could re-identify participants are included;
The remaining demographic variables are limited to general information, including age range, sex, height range, and weight range, which are not sufficient to identify individuals when combined;
The provided dataset does not contain any video, audio, or image data that could be used to identify individuals;
Data were reviewed by the research team to confirm that no indirect identifiers or combinations of variables pose a risk of re-identification.

These measures comply with the data sharing standards required by Dryad and ensure that the dataset is suitable for open public release.

This study recruited 23 healthy adults aged between 21 and 30 years old (9 males, 14 females). All participants had no neuromuscular diseases or skeletal injuries, and without any hearing or vision impairment. In the experiment, participants were required to walk back and forth on a 7m × 1m mat under a single-task condition and three non-standardized dual-task conditions. This study utilized eight MOCAP cameras (NOKOV MARS 2H HD cameras, 100Hz, resolution 2048 × 1088), one depth camera (Microsoft Kinect V2.0, 30Hz, resolution 512 × 424), and one inertial sensor (Witmotion BWT901BLECL5.0C, 100Hz) for data collection.

For the walking data from each task recorded by MOCAP, 22 reflective markers were firstly labeled using MOCAP's software and the 3D trajectories of the markers were then exported. The data collected by IMU were exported using IMU's software, and the data collected by Kinect were exported using its proprietary SDK. The data collected by the three sensors were all exported in .csv format. After that, for partially missing values of certain markers in the three-dimensional trajectories exported by MOCAP, a relational interpolation method based on the movement of surrounding markers was applied using Anaconda3. To remove high-frequency noise, the data collected by MOCAP and IMU were smoothed using a Butterworth low-pass filter with a cutoff frequency of 7.5 Hz. The noise and fluctuations of the data collected by Kinect were reduced using a moving average with a window size of 3. All the data were then segmented using six manually recorded timestamps from the experiment. To minimize the impact of gait initiation on walking speed, the final gait dataset excluded the initial and final 2 m of walking. The data collected by MOCAP and IMU were segmented into 5m back-and-forth data, 2 m back-and-forth data, and turning data, while the data collected by Kinect was segmented into 2 m back-and-forth data and turning data. Attribute to the Kinect's erroneous identification of left and right joints as reversed during the backward phase, we adjusted the coordinates by swapping the left and right joints during this phase. After data preprocessing, we extracted a total of nine spatio-temporal gait parameters and fourteen kinematic parameters.