Abstract: A novel direction finding system based on radial-basis function neural network is proposed to solve the problem of direction of arrival estimation for an array with system errors(mutual coupling and mismatching).Through calibration of input data with direct data domain algorithm,the correct centers of radial-basis function can be obtained.Direction of arrival can be estimated without improving the radial-basis function neural network DOA estimation system.In order to reduce the dimension of input vectors,the symmetry property in the correlation matrix is utilized.Only the upper triangular half of the correlation matrix is used in processing.The specific steps are presented.The simulation results demonstrated that this DOA estimation system has very high precision.