This paper establishes a clustering model using Chinese characters in text document as its variable based on fuzzy clustering theory.According to human visual system,it classifies the Chinese characters in text documents by their stroke number,appearance frequency and distribution uniformity.Then it chooses the Chinese characters with less stroke number,higher frequency and higher distribution uniformity to be embedded.Experimental results demonstrate the effectiveness of the algorithm against partial deleting,adding and format adjusting of text.The algorithm has good concealment.