Abstract: According to noise problems before LDPC decoding,we propose LDPC decoding algorithm,which based on compressive sensing reconstruction for denoise.First of all,at the receiving end,we use CS algorithm observing and recovering the received signal of system,to eliminate the noise in the process of information channel transmission,and then we use the restoring signal as a received signal directly into the LDPC decoder.The simulation calculation shows that the improved algorithm can effectively reduce the effects of noise,reduce the LDPC error rate and improve the decoding performance of the system,when the code length is 512,the error rate can be reduced to 10-5.And the error rate is influenced by the sparsity,the transmission rate and CS reconstruction algorithm.Comparing to four kinds of CS greedy algorithm reconstruction,SP algorithms get better.