Abstract: A time-domain deconvolution method of bearing estimation combining multi-channel high-order statistics is proposed for single-vector hydrophone. The second-order statistics of multiple channels of a vector hydrophone is used to construct a beam pattern with narrow main lobe and invariant translation, and the deconvolution algorithm is used to realize the bearing estimation. The proposed method has much narrow main lobe width, higher resolution and robustness. Simulation experiments and sea trial data show that the proposed method outperforms other methods such as multiple signal classification (MUSIC) in terms of dual-target beam resolution and bearing estimation accuracy under non-ideal conditions where the response of each channel of the vector hydrophone is inconsistent, and the dual-target beam resolution is improved by about 40%.