Aimed at the problems that by the traditional recognition method the recognition rate is low and the feature extraction is difficult under condition of low signal to noise ratio, an automatic classification and recognition system based on short time Fourier transform (STFT) and stacked denoising AutoEncoder (sDAE) is proposed. Firstly, the radar emitter signal is changed by the short time Fourier transform, then a series of preprocessing is carried out to the time frequency image. The processed image is input into the stacked denoising autoencoder, and the extracted features are input into the softmax classifier to complete the classification recognition. The simulation shows that the recognition rate can reach more than 80% when the system is in SNR=-10 dB. In the case of low SNR, the recognition rate is obviously better than that by the traditional recognition method.