The invention relates to an interference noise matrix reconstitution-based self-adaptive wave beam forming method and belongs to the field of array signal processing. The method comprises the steps of firstly establishing the signal model of a minimum variance distortionless response wave beam forming problem, then under the condition that the angle range of an expected signal wave arrival direction is known, utilizing a multiple signal sorting spatial spectrum to reconstitute an interference noise covariance matrix in a region without containing expected signals, based on the matrix, solving the estimation value of a guide vector of the real expected signal by using the maximizing of an array output power and the constrain condition that the estimation value of the guide vector of the expected signal is prevented from being converged to the guide vector of any interference or the linear combination of the interference. According to the method, a simulation result shows that when random pointing errors and local scattering of the expected signal and a interference source exist, the output signal to interference plus noise power ratio in a very large input signal-to-noise ratio range is still close to a theoretical value and is superior to that of other self-adaptive wave bean forming methods.