To address the intricate challenges of estimating parameters in underwater acoustic channels, this paper introduces an l0 norm constrained sparse Bayesian learning method. The algorithm gives a prior distribution of the channel to be estimated. The algorithm provides the maximum posterior estimate based on Bayesian principle. This algorithm effectively enhances the sparsity of the estimated channel by integrating penalty terms into the cost function. Experimental results demonstrate the effectiveness of the l0 norm constrained sparse Bayesian method. The results show a significant improvement in estimation accuracy with increasing SNR. We compare this algorithm with the least squares algorithm and the orthogonal matching pursuit algorithm. Under low SNR conditions, the estimation accuracy of the l0 norm constrained sparse Bayesian method closely resembles that of the least squares method. Conversely, under high SNR conditions, the algorithm outperforms the least squares method, approaching the accuracy levels achieved by the orthogonal matching pursuit algorithm. Experiments show that the l0 norm constrained sparse Bayesian algorithm can effectively estimate the underwater acoustic channel.