Blind recognition of convolutional codes is the basis for recognition of certain high performance codes including concatenated and Turbo Code.It requires that the recognition methods for convolutional codes should have strong robustness against channel noise.The key to such purpose is to make use of the received soft information.Firstly

this paper gives a probabilistic analysis about the reason why the existing methods using soft information performs no better than the method based on hard information.The reason is that the candidate solution vectors of low Hamming weights seriously deteriorate the correct recognition probability.Then

a solution based on least-square cost function is proposed for this problem.Theoretical analysis proves that the impact of low Hamming weights can be effectively reduced.Finally

the theoretical results are verified by simulation experiments.Both the theory and the simulations show that

for blind recognition of convolutional codes

the proposed method improves the robustness against noise by about 1dB.