4, our approach utilizes only a few non-testing spike sequences with labeled 3D poses as reference sequences. These sequences are aligned to generate a spike pose model. This model captures the relative position and distribution at the spatial level, as well as the movement trend and ...