ing at the interference problem introduced by satellite multipath signals, a blind source separation model of satellite multipath interference signals is established. The joint approximate diagonalization of eigenmatrix (JADE) algorithm based on dimensionality reduction Householder transformation is used to perform the joint approximate diagonalization of the fourth-order cumulants of received mixed signals, so as to extract multipath signals. Experimental results show that the algorithm can well separate the multipath signals with spectrum completely overlapped and the interference signals with spectrum partially overlapped. In terms of extracting multipath signals, the performance of this algorithm is superior to fast independent component analysis (FastICA) algorithm . In terms of running time, the average time of FastICA algorithm running 1000 times is about 6 times that of the Cardoso Givens-JADE(CG-JADE) algorithm, and the average running time of this algorithm is 0.0024 s less than that of GG-JADE algorithm, which proves its effectiveness and rapidity.