Towards a multimodal human activity dataset for healthcare

Human activity recognition (HAR) based on wearable devices has become a hot topic due to the wide adoption of smartphones and smart bands. In this paper, we propose a new dataset, MMC-PCL-Activity, for wearable device-based HAR. It contains data of accelerometers, gyroscopes, heart rates, steps, GPS, weather information, mobile APP usage, and images collected from 14 participants performing 16 different types of daily activities. Besides the activity annotations, labels of physical health status and mental health status are also provided. We demonstrate the importance of multimodal fusion in activity recognition and provide baselines for more researchers using this dataset.

Automated summary

This paper introduces a new dataset for wearable device-based human activity recognition (HAR), which includes multiple sensor data and labels of participants' health status. It demonstrates the importance of multimodal fusion in activity recognition and provides baselines for further research using this dataset.