The Data is from the Ubicomp 2023 paper entitled "mmStress: Distilling Human Stress from Daily Activities via Contact-less Millimeter-wave Sensing".
Paper DOI: https://doi.org/10.1145/3610926
Please cite the paper when you use our data.
Data is include two folders named "Cleaned_mmWave_Activity_Data" and "Stress_Value_from_Garmin_Watch", which represent activity data from mmWave radar and stress label from Garmin watch respectively.
In this work, 10 volunteers participated in the long-term deployment experiment, and 8 volunteers returned valid data (data the other two volunteers is severely missed due to uncontrollable factors). After obtaining permission from the 8 volunteers, we make these data public.
Thanks to these 10 volunteers for their contributions to this work.
As mentioned in the paper, we install the mmWave radar on a wall 2 meters above the ground, with a downward inclination angle of 15 °.
In the folder named "Cleaned_mmWave_Activity_Data", 8 volunteers activity data is saved as CSV format.
CSV contains 5 attributes:
devid: Radar ID in this work;
mytime: Timestamp accurate to milliseconds;
pointcloud: All point clouds present in this frame, represented in JSON format
{“pointcloud”: [[ x1,y1,z1,range1,angle1,elev1,doppler1], [ x2,y2,z2,range2,angle2,elev2,doppler2],… [x(n),y(n),z(n),range(n),angle(n),elev(n),doppler(n)]]}
trackedobj：Using the method in the paper to generate human activity trajectories based on point clouds, represented in JSON format
{“trackedObj”:[[ x1,y1,z1,vel_x1,vel_y1,vel_z1,acc_x1,acc_y1,acc_z1,Target_ID1], [ x2,y2,z2,vel_x2,vel_y2,vel_z2, acc_x2,acc_y2,acc_z2,Target_ID2],… [x(n),y(n),z(n),vel_x(n),vel_y(n),vel_z(n), acc_x(n),acc_y(n),acc_z(n),Target_ID(n)]]}
After data cleaning, there is only one trajectory kept.
pointcloudassociatedid: The trajectory ID corresponding to the point cloud is used during data cleaning and is now useless, represented in JSON format
{“pointCloudAssociatedID”:[pointcloudID1,pointcloudID2,…pointcloudID(n)]}
As mentioned in the paper, we use stress class generated by Garmin watch as the ground truth. The stress value generated every 3 minutes. In the folder named "Stress_Value_from_Garmin_Watch", we save everyday stress value changes for 8 volunteers using CSV format. Each CSV file includes two attributes：
StressTime：Timestamp accurate to milliseconds;
StressValue: Stress value generated by Garmin Watch.
Please refer to the paper for details on the generation of stress levels and the selection of time windows.
Since the data file is too large and exceeds the github LFS quota, we migrated the data to Google Drive: https://drive.google.com/drive/folders/1u0OUbtKPbKi0laWQWEHVeQvjp-wl4Sli?usp=drive_link , you can access our data through this URL.