Abstract: In order to achieve reasonable and reliable results of directional spectrum used in wave measurement of Acoustic Doppler Current Profiler (ADCP), three practical inversion algorithms based on array measurement are analyzed on performance in theory and used in translating water surface elevation collected by ADCP (both computer simulated and experimental data included) into directional wave spectra for comparison. It is shown that Bayesian Directional spectrum estimation Method (BDM) is most accurate in directional resolution; Maximum Likelihood Method (MLM) reaches sufficient stability of solution; and with long wavelengths, Extended Eigen- Vector (EEV) method can obtain most specific and sharp peaks of directional spectra. Therefore, MLM is suitable for real-time processing of wave data, and BDM could be used when higher accuracy is demanded.