A novel method for segmenting simplified skeletons of 3D shapes is presented. The so-called simplified Y-network is computed, defining boundaries between 2D sheets of the simplified 3D skeleton, which we take as our skeleton segments. We compute the simplified Y-network using a robust importance measure which has been proved useful for simplifying complex 3D skeleton manifolds. We present a voxel-based algorithm and show results on complex realworld objects, including ones containing large amounts of boundary noise.