Experimental data

Unlike in our NIPS 2017, now you can use publicly available experimental data to bootstrap the algorithm. Example running and walking kinematics can be found in many publicly available papers. For example we refer to Schwartz, M. et al. (2008) (citation details below).

You can find there joint angles and EMG signals. Data is represented as a function of time in a gait cycle (one step) at different speeds.

You can use it for supervised learning for bootstrapping your models (where for given kinematics you predict muscle activity).

@article{schwartz2008effect,
  title={The effect of walking speed on the gait of typically developing children},
  author={Schwartz, Michael H and Rozumalski, Adam and Trost, Joyce P},
  journal={Journal of biomechanics},
  volume={41},
  number={8},
  pages={1639--1650},
  year={2008},
  publisher={Elsevier}
}

Another great dataset includes running data from

@article{hamner2013muscle,
  title={Muscle contributions to fore-aft and vertical body mass center accelerations over a range of running speeds},
  author={Hamner, Samuel R and Delp, Scott L},
  journal={Journal of biomechanics},
  volume={46},
  number={4},
  pages={780--787},
  year={2013},
  publisher={Elsevier}
}

Note that in also contains inverse dynamics (predicted muscle activations from experiments) as well as EMG signals!