Abstract: A new approach is presented for blind estimation of linear block codes parameters with noisy data. Firstly, the proposed method exploits the statistical properties of the dual codes and Walsh-Hadamard transform to determine whether a code is belong to dual code or not. Meanwhile, from the principle of 3 standard deviation, a theoretical threshold to distinguish dual codes is induced. Then the code length and synchronization are estimated when the normalized dual space dimension reaches the maximum. Finally, the parity-check matrix is recovered by the dual codes. Computer simulation results show that the proposed method can provide good performance even when the bit error rate is 0.03.