That's good then, maybe you can use a separate point cloud to keep track of the point velocities. In its modelling region, add a particle at each of the mesh's vertex positions. In its simulation region, get the mesh's current point positions, subtract the cloud's point positions (you'll need to use Switch Context for this), use the length to set the attribute that you are using to filter the other cloud's emission (again, use Switch Context), and finally set the particles' point positions to the mesh's point postions to be ready for the next frame.