### Summary

This paper extends the work done in the registration curves paper by applying it to an entire library of motions for the purpose of automatic identification of similar motions and generation of the corresponding curves. This method provides a way of determining an interconnecting web of motions that are similar to each other allowing users to efficiently search the library for related motions and interpolate between them if necessary to create new motions.

### Problem

Large libraries of motion data contain many similar motions that could be used to generate new motions. However, locating such motions is prohibitively time consuming.

### Methods

1. The distance metric and registration curve ideas from the previous paper are used to locate similar motions and generate the appropriate curves for interpolating between them.

2. A multi-resolution approach is used to speed up the process of identifying possibly similar motions in the library.

3. A match web is used to connect similar motions allowing different motions to be blended together via these connections.

### Contributions

1. Methods for constructing a searchable database of similar motions from which interpolation can be used to generate motions.

2. Generation of parameterized motions from sets of similar motions.