Abstract: Compressed sensing is a novel signal processing theory that it introduces a novel way of acquiring compressible signals, the test times of existing sparsity trial and error algorithms were always large. The novel algorithm, blind sparsity adaptive matching pursuit (BSAMP) was proposed, could recover the original signal fast in the case of unknown sparsity. Firstly, the range of sparsity was determined, and each time half of values in current range were eliminated by trial and error test. Secondly, the number of atoms was twice the sparsity, which was united with the set of signal approximation support (got by last iteration) and then reconstructed the signal by solving least-squares problems. Last but not least, the least-squares approximation was pruned by weakly matching for next iteration. The results of simulation show that the novel algorithm can reconstruct signal faster and get larger recovery probability than other similar algorithms in the same conditions.