In order to filtering out background noise effectively in on-site electromagnetic emission test

virtual dark chamber test method based on compressive sensing（ CS） is proposed to solve the problem of failure of algorithm for multiple signal classification（ MUSIC） when correlation signals exist in environment. Using theory based on CS to achieve directions of arrival（ DOA） of radiation signal of equipment under test（ EUT） and other interference signal in environment

and using minimum variance distortionless response（ MVDR） bream-forming algorithm for beamforming

so as to effectively suppress interference signal. This method can not only be applied in the presence of coherent signals

but also has more flexible applications without restrictions in array structure.Simulation results prove that the method can suppress interference noise effectively without distortion of character of EUT.