Isolated sign language recognition is a long-standing research problem. Existing methods consider inclusively ambiguous data to represent a sign, and ignore the fact that only a few key in-formation can represent the sign efficiently since most information are redundant. Furthermore, inclusion of redundant information may result in inefficiency and difficulty in modeling the long-term dependency for sign language recognition. This letter delivers a novel sequence-to-sequence learning method based on key-frame centered clips (KCCs) for Chinese sign lan-guage recognition. Different from conventional methods, only key information is considered to represent a sign significantly. The frames-to-word task is transformed into a KCCs-to-subwords task successfully, to allow for different at-tention in the input data. The empirical results of the proposed method outperform significantly the state-of-the-art sign lan-guage recognition systems on our dataset containing 310 Chi-nese sign language words.