Abstract: An algorithm of distributed signal recognition is presented to improve the spectrum sensing ability of sensor network. The proposed method extracts highorder cumulants and spectrum characters of modulated signal in the single sensor node. The fusion center assigns the weighting values to the sensor nodes according to SNR estimation, and fuses the character parameters of sensor nodes on the character layer to fulfill the signal recognition. Simulations show that when the SNR is no less than -4 dB, for nine kinds of modulated signals, the proposed algorithm can get the correct recognition rate of 90% in AWGN channel. And the algorithm can still keep the high sensing ability in the case which some sensor nodes are in worse environment.