MPJA-HAD: A Multi-Position Joint Angle Dataset for Human Activity Recognition Using Wearable Sensors
Hongmei Yang, Xiaoxu Wen, Yingrui Geng, Yan Wang, Xin Wang, Chenggang Lu
Abstract
The Inertial Measurement Units (IMUs) have been widely used in human activity recognition (HAR) for data acquisition. However, most publicly available datasets based on IMUs only involve data from few body parts and are relatively homogeneous. Using data from few body parts can be limited in certain specific or complex activity recognition tasks. Hence, we created a new HAR dataset named Multi-Position Joint Angle Human Activity Dataset (MPJA-HAD). Other publicly available datasets only contain the raw inertial sensor readings. The MPJA-HAD dataset also provides joint angle changes from each of 15 body positions. Joint angles directly relate to human activities performing and experimental results show the competitiveness of the created dataset in HAR tasks.