针对交流毫米波辐射信号丢失部分低频分量的缺陷，使用RVM算法在交流毫米波辐射信号的DCT变换域补偿出丢失的成分，再通过IDCT变换即可获得补偿的交流毫米波辐射信号。RVM使用贝叶斯方法推理，推广能力好，与支持向量机相比解更为稀疏，这些保证了RVM算法具有良好的预测能力。实验结果也显示本方法可以获得良好的补偿效果。

Aimed at the shortcoming that the loss of lowfrequency information of alternating current millimeterwave radiometer signal, relevance vector machine (RVM) algorithm is used to compensate the lost component in discrete cosine transform (DCT) domain, and through inverse discrete cosine transform (IDCT) we can receive the compensated signal. RVM exploits Bayesian learning framework, which has dramatically fewer kernel functions than comparative support vector machine. So that accurate prediction models can be acquired. Experimental results also show that this method can obtain good compensation effect.