Abstract: Ship noise is the main factor that affects the quality of underwater voice communication. In order to improve the SNR and the quality of the single side band voice signal, an intensive study is made by using the multi-channel adaptive enhancement algorithm. The adaptive signal enhancement is adopted in this algorithm, which has the ad-vantages of low computational complexity, easy real-time implementation and remarkable denoising effect. In addition, the information received by different transducers can be effectively utilized. In this paper, the algorithm is analyzed theoretically, and the simulation is carried out with the sea trial data. Simulation results show that the proposed algorithm can effectively suppress the sharp noise in low SNR and non-stationary noise. And when the system is convergent, the noise tends to be whitened. Besides, the output SNR is obviously improved, the quality of voice is improved significantly, and the performance is much better than that of the traditional signal coherent superposition.