Abstract: A method of motion data processing based on sampling-Isomap is proposed. By computing the distance matrix of sample frames, the sample embeddings eigenvectors can be obtained to approximate the eigenvectors of the original embeddings, and then the non-sample frames can be projected in this embeddings to approach the manifold embeddings of non-sample frames. Experiments proved that approximate embeddings of motion data computed by sampling-Isomap were average 10 times faster than by Isomap, while 10% frame samples were selected. Motion analysis can be performed efficiently even with high-dimensionality of motion capture data.