In this paper, based on the Swin Transformer neural network we propose a method for the recognition of modulation type of the Low Probability of Intercept (LPI) radar signal. We firstly perform the time-frequency transformation, that is to say, transform the one-dimensional time-domain signal into a two-dimensional time-frequency image by using the smooth pseudo Wigner-Ville distribution. We then use the Swin Transformer neural network to extract the desired features from the image and identify the modulation type. Numerical simulation shows that this method processes strong anti-noise ability and high recognition accuracy even under the condition of low signal-to-noise ratio. Meanwhile, it is also shown that this method has strong adaptability to small samples.