Abstract: Considering the effects of underwater acoustic channel on underwater acoustic communications,the capacity of underwater acoustic communication system is analyzed,and an underwater acoustic channel capacity assessment algorithm is studied.A piecewise symbol-variance transfer function model based on the a posteriori Gaussian distribution is proposed via the statistical analysis of the a posteriori symbols which are the output of equalizers under different conditions.The Maxwell-Boltzmann distributions are used to perform probabilistic shaping on the output symbol of equalizer.Probabilistic shaping technology is introduced to analyze and assess capacity of the actual underwater acoustic communication system.Finally,the capacities with different modulation modes are obtained by analyzing the multi-path channel data of pool and lake experiments.The results show that the maximum improvement of the channel capacity by using probabilistic shaping technology over the conventional equiprobability mapping is 0.5 bits/symbol,in the specific Signal-to-Noise Ratio(SNR) region.