@inproceedings{bauer2014interactive,
  title={Interactive Visualization of Muscle Activity During Limb Movements: Towards Enhanced Anatomy Learning},
  author={Bauer, Armelle and Paclet, Florent and Cahouet, Violaine and Dicko, Ali Hamadi and Palombi, Olivier and Faure, Fran{\c{c}}ois and Troccaz, Jocelyne and others},
  booktitle={Eurographics Workshop on Visual Computing for Biology and Medicine,(VCBM) 2014},
  year={2014}
}

Abstract
We propose a framework to investigate a new way to learn musculoskeletal anatomical kinetics using interactive motion capture and visualization. It can be used to facilitate the learning of anatomy by medicine and sports students, and for the general public to discover human anatomy in action. We illustrate our approach using the example of knee flexion and extension by visualizing the knee muscle activation prediction with agonist and antagonist co-contraction. Muscle activation data for specified movements is first measured during a preliminary phase. The user is then tracked in real time, and its motion is analyzed to recognize the motion being performed. This is used to efficiently evaluate muscle activation by interpolating the activation data stored in tables. The visual feedback consists of a user-specific 3D avatar created by deforming a reference model and animated using the tracking. Muscle activation is visualized using colored lines of action or 3D meshes. This work was made possible by the collaboration of three complementary labs specialized in computer-aided medical intervention, computer graphics and biomechanics

ESTO ES MUY PARECIDO A LO QUE QUEREMOS HACER NOSOTROS! 
HABRIA QUE VER EN QUE NOS DIFERENCIAMOS Y POR QUE PODEMOS DAR MEJORES PROPPUESTAS
Basicamente graban actividad de EMG y posiciones y velocidades angulares de flexion de piernas en una etapa experimental, de un solo sujeto.
Luego lo que hacen es escalar el RMS del EMG entre 0 y 1 (siendo 1 el maximo valor medido). Eso como que lo almacenana en una base de datos.
Despues para nuevos sujetos mejoran el tracking de kinect y obtienen las mismas variables kinemáticas y por interpolacion con los datos experimentales de la base de datos devuelven la activacion que le corresponderia
Finalmente esto lo muestran en un modelo 3D sobre un avatar aumentando con información grafica de los musculos coloreados en base al nivel de activación recuperado

The authors of \cite{bauer2014interactive} propose a framework of Augmented reality to visualize muscle level of activity in 3D of lower limb movements.
In a fisrst experimental step, they capture kinematic information (with a motion capture system) and EMG of 10 surface muscles of an experimental subject performing a series of lower limb movements and store this information in a database. The EMG signal is normalized to the maximal captured value and the activation levels of deeper muscles is obtained by models.
In a second step when a subject use the system, they track kinematic information using an immproved version of Kinect skeleton tracking algorithm. Performing interpolation with the experimentally captured data they retrieve the activation levels corresponding to the new sensed movement.
Finally the activation leveles is shown in a colour scale by 3d Models of the muscles overlapped with an avatar to show it in an friendly user interface.