The polarized antenna is an attractive technology in many applications, such as the space communication with polarization diversity, frequency division communication with extra isolation level, and anti-interference communication, etc. However, the polarization effect often becomes attenuated after the signal’s propagation progress, which leads to the channel depolarization, or the mismatch of the polarization direction between the transmitter and receiver. A framework of machine learning assisted signal detection method is proposed, in which the compressive sensing (CS) is leveraged for the restore of signal, and the Artificial Neural Network (ANN) is used to identify the serious depolarization effect to ensure the data recovering effect. The results of the simulation experiment verify that this method can significantly reduce the loss of the signal information caused by the depolarization effect.