Compressive sensing (CS) is an emerging field based on the revelation that a small collection of linear projections of a sparse signal contains enough information for stable, sub-Nyquist signal acquisition. Ultra-wideband (UWB) noise radar is one of the novel techniques which is widely used in various applications such as emergency rescues and military operations. One challenging problem in UWB noise radar operation is that a huge amount of data will be received which requires tremendous storage space. Compressive sensing could easily handle his problem since it captures all the information we need from far fewer samples. In his paper, we propose a novel amplitude based compressive sensing algorithm to compress data without any knowledge in advance. Simulation results indicate that only 1/5 of original measurements are sufficient to recover original data, which also achieves a higher compression ratio than the conventional compressive sensing.