We offered following four benchmark datasets as part of the project:
HumCareHome Dataset
Description: The dataset contains recordings of 11 everyday activities, including sitting, standing, lying, jogging, cleaning a table, and more. The activities were performed in a controlled environment of a home by 12 participants, each for a duration of 3 minutes. The dataset captures a mix of static and dynamic behaviors, providing a rich resource for modeling real-world human activity scenarios.
Data Collection: Motion data was simultaneously collected using three wearable devices: a smartphone, a smartwatch, and smart glasses. Each device utilized inertial measurement unit (IMU) sensors, specifically accelerometers, gyroscopes, and magnetometers. This ensures comprehensive motion capture from multiple body segments (head, upper body, and lower body) for detailed and holistic activity profiling.
@article{ashfaq2025hcmma,
title={HCMMA-Net: A Hybrid Convolutional Multi-Modal Attention Network for Human Activity Recognition in Smart Homes Using Wearable Sensor Data},
author={Ashfaq, Nazish and Aziz, Zeeshan and Khan, Muhammad Hassan and Nisar, Muhammad Adeel and Khalid, Adnan},
journal={IEEE Access},
year={2025},
publisher={IEEE}
}
Description: The HumcareFall dataset captures 8 types of fall-related activities performed by 61 participants. Each activity was recorded for 5 seconds, providing focused, high-resolution segments essential for developing and evaluating fall detection systems.
Data Collection: Data was collected using two wearable devices: a smartwatch and a smartphone, equipped with inertial measurement unit (IMU) sensors, including accelerometers and gyroscopes. This setup enables accurate motion tracking from multiple body positions, supporting robust modeling of fall events.
Description: The HumcareADL dataset contains recordings of 30 activities of daily living (ADL), performed by 60 participants. Each activity was recorded for 5 seconds, capturing a diverse range of human behaviors crucial for activity recognition and assistive technology development.
Data Collection: Motion data was gathered simultaneously using three wearable devices: a smartwatch, smart glasses, and a smartphone, all equipped with inertial measurement unit (IMU) sensors, including accelerometers, gyroscopes, and magnetometers. This multi-device setup ensures detailed motion profiling across different body regions (head, upper body, and lower body) for comprehensive activity analysis.
Description: The HumCareAF dataset provides a comprehensive collection of human activity data, combining fall-related events and activities of daily living (ADL). It includes recordings from 87 participants, each performing a diverse set of 8 fall-related activities and 30 ADL activities. Each activity was captured over 5-second intervals, offering short, high-resolution time-series segments essential for training and evaluating advanced human activity recognition and fall detection systems.
Data Collection: The HumCareAF dataset was collected using a combination of wearable smart devices equipped with inertial measurement unit (IMU) sensors, including accelerometers, gyroscopes, and magnetometers (where applicable). For fall-related activities, motion data was captured using two devices: a smartwatch worn on the wrist and a smartphone carried in the pocket. This setup enabled accurate motion tracking from the upper and lower body, essential for identifying dynamic fall patterns. In contrast, activities of daily living (ADL) were recorded using a more comprehensive configuration involving three devices: a smartwatch, a smartphone, and smart glasses. This tri-device setup provided synchronized motion data from the head, upper body, and lower body, allowing for fine-grained analysis of everyday human activities. By tailoring the device configuration to the context of the activity, the dataset ensures high-quality, context-aware motion profiling suitable for robust modeling of both fall detection and activity recognition tasks.
URL: https://drive.google.com/drive/folders/1hILElqfhMFgpq5300YUBReSV3whhoUE8?usp=sharing Paper: Please cite the following paper to use this dataset.