The USC Human Activity Dataset (USC-HAD)

Dataset Description

The intention of the USC Human Activity Dataset (USC-HAD) is to serve as a benchmark for human activity recognition algorithm comparison and make research efforts based on it repeatable and extendible by researchers in ubiquitous computing and machine learning communities. The dataset is specifically designed to include the most basic and common low-level human activities in daily life from a large and diverse group of human subjects. Our own focus is on healthcare related applications such as physical fitness monitoring and elder care, but the activities in the dataset are applicable to many scenarios. The activity data is captured by a high-performance factory-calibrated inertial sensing device called MotionNode and is well-annotated. In total, we have included 12 activities and collected data from 14 subjects. Here is a list of summarized properties of the dataset:


Download the Dataset

The dataset (in .zip format) can be downloaded HERE (File Size: 42.5MB). The data is stored and organized in MATLAB format. Please read the readme.txt in the archive for details on accessing the data.


Citation

Mi Zhang and Alexander A. Sawchuk, "USC-HAD: A Daily Activity Dataset for Ubiquitous Activity Recognition Using Wearable Sensors", ACM International Conference on Ubiquitous Computing (UbiComp) Workshop on Situation, Activity and Goal Awareness (SAGAware), Pittsburgh, Pennsylvania, USA, September 2012.  [ PDF ] [ BibTex ]


Important Note

The USC-HAD dataset is the property of the Signal and Image Processing Institute of Ming Hsieh Department of Electrical Engineering at USC. This dataset can only be used for research and non-commercial purposes. Please cite our paper if you publish results based on this dataset. If you have any questions or suggestions about the dataset, feel free to email me.