[Algorithms] Representing Animation Key Frame with Quaternion + Translation Vector

[Lim S. C. <si...@ga...>]

Since a quaternion is compact compared to a 4x4 homogeneous matrix, I am considering representing our animation key frame data with pairs of orientation quaternion and translation vector. Naively, this seems to save me 56.25% of memory space and that sure comes in handy when we have a lot of animation data.
   
  Just wondering if anyone has done this before and whether it is really better in terms of performance and accuracy.
   
  One problem I still haven't solved is concatenating my quaternion-vector key frames without converting them into matrices first. I know I can concatenate quaternion by taking their cross product, but I do not know how to do so for the translation vector. I will be much obliged if someone could show me a few hints or pointers.
   
  Cheers!