Aiming at the radars classification and recognition of the interference signal under complicated electromagnetic environmentwe studied the Choi-Williams Distribution (CWD) time-frequency images of the RF noise interferencenoise-amplitude-modulation interferencenoise-frequency-modulation interferencerange gate pull off interference at a constant speedand speed gate pull off interference.The AlexNet convolution of deep learning neural network model was adopted to automatically extract the image features of detailsso as to realize the classification and recognition of radar intereference signal.The simulation results showed that:1) The recognition rate of the network increases rapidly with the increase of the interference-to-noise ratio in the range of -10 dB to 0 dBand the recognition rate is basically close to 100% when the interference-to-noise ratio is above 0 dB;and 2) In the full range of interference-to-noise ratiothe networks recognition accuracy rate is 99.25%and the recognition effect is good.