### Summary

This paper discusses a solution to the issue of an artifact known as a "footskate." Due to imperfect motion capture data, preprocessing, or other causes, a foot may not properly plant on the ground. The authors of this paper propose a solution to this problem which automatically identifies footplants, but allows a human to override it and adjust the constraints. The algorithm then modifies the motion capture data to adhere to the constraints, smoothly blending the motion between constrained frames as well as between constrained and unconstrained frames. The algorithm presented.

### Problem

Footskate artifacts look unnatural, yet are very common in motion capture data sets either as a result of bad data or artifacts introduced by post processing the data.

### Methods Used

1. Position and orient the feet in constrained frames to match the constraints. The positions and orientations are blended across the animation.

2. Use a modified version of a single-limb IK solver to position the rest of the lower body. The modified version incorporates "knee-damping" which lengthens the leg in order to prevent "knee pop" which is the large change in the angle of the knee needed to extend the leg when the leg is near full extension.

### Key Ideas

1. The introduction of "knee damping" in order to minimize the effects of knee pop by allowing the legs to slightly increase in length.

### Questions/Comments

I find it curious that the human mind would find the lengthening of the leg to be less disturbing than rotating the knee quickly. However, it does make sense. Just another one of those "more real than real" things.