Abstract: Impulsive noise represents a prevalent interference in underwater acoustic channels, and when impulsive noise and orthogonal frequency division multiplexing (OFDM) signal peaks are superimposed on each other, the detection and suppression of impulsive noise will become more difficult, which leads to severe performance degradation of the underwater acoustic communication system. To address this problem, an underwater acoustic impulsive noise suppression method based on adaptive median filter is proposed. Firstly, the pilot subcarrier information is utilized to obtain the background noise pre-estimation value, from which the impulsive noise emergence location is detected and marked. Then, the impulsive noise is suppressed by the adaptive median filter to change the distribution characteristics of the background noise, which improves the performance of the sparse Bayesian learning algorithm channel estimation algorithm. Simulation results based on the experimental impulsive noise and the communication data of the 13th Arctic Research Expedition show that the proposed method can effectively suppress the impulsive noise component in the background noise, which has better bit error rate (BER) performance relative to the traditional method. Under the −10~20 dB generalized signal-to-noise ratio, the BER of proposed method can be reduced by 5.42% and 11.15% on average, respectively.