Abstract: In this paper, a novel recognition method of single-hand static words of Chinese Sign Language based on fuzzy-neuro network is introduced. First, the fuzzy reasoning rules and the network structure are established using empirical knowledge. Then the memloership function parameters for each fuzzy subset are obtained by learning. For the learning process, a new kind of empirical risk function is proposed which is differentiable and can be minimized hy gradient descent strategy. This method is compared with others through experiments and its validity and reliability are confirmed.