A rigid map neural network for anatomical joint constraint modelling

Accurate individual anatomical joint models are becoming increasingly important for both realistic animation and diagnostic medical applications. A number of recent approaches have exploited unit quaternions to eliminate singularities when modelling orientations between limbs at a joint. This has resulted in the development of unit quaternion based joint constraint validation and correction methods. A number of machine learning approaches have been applied to this problem. Recent work has demonstrated the use of Kohonen's Self Organizing Maps (SOMs) to model regular conical constraints on the orientation of the limb. In this paper we investigate a derivative of the SOM, the Rigid Map, applied in the same context.