CEI BioTIC

Available datasets

Datasets used to test different techniques for human behaviour learning and recognition:

  • This dataset contains 7 actions (walk, twist left, twist right, bend up, bend down, sit down, stand up), performed by 17 participants (10 repetitions by participant). If you use this dataset, you may find useful to cite the following article for comparison of results: O. Leon, M.P. Cuellar, M. Delgado, A Nivel Approach for Real-Time Learning and Recognition of Human Activities, International Journal of Ambient Intelligence and Smart Environments (JAISE), submitted, 2014 (submitted)

    The table Behaviour contains information about the name of the activity, the ID of the user that performed it, number of frames, and date of recording. On the other hand, the table Skeleton contains the frames for each activity in Behaviour, ordered by frame. The remaining of the attributes are the joint position (3-D) of each joint obtained from Kinect

    .
  • This dataset contains 5 diagnosis exercises in physical therapy described in the paper published at ELEMENT-2014, performed by 10 participants (10 repetitions by participant). If you use this dataset, you may find useful to cite the following article for comparison of results: M.P.Cuellar, M. Ros, M.J. Martin-Bautista, Y. Le Borgne, G. Bontempi, An approach for the evaluation of human activities in physical therapy scenarios, Int. Conf. ELEMENT-2014 (submitted)

    The file FilesDescription.txt contains information about the content, but you are encouraged to contact with us for any concerns

    .