A novel logical-based similarity motion retrieval method is proposed for meeting content-based motion retrieval. Firstly, the self-organizing mapping (SOM) algorithm is used to find the best matching units (BMUs) in accordance with the order data specially designed for human body motion, and then, the encoding method is implemented to establish the motion sequences descriptor based on the symbolic aggregate approximation algorithm. Further more, the SAX-based descriptor is used to build the index mechanism and matching. Finally, the proposed method is used to retrieve motion data. The experiment results show that the proposed algorithm is effective.