Blind Recognition Algorithm for Scrambled Channel Encoder Based on the Features of Signal Matrix and Layered Neural Network

Blind recognition is of great significance in non-cooperative communication research. In practical communication scenarios, multistage signal processing schemes are usually used, including scrambling, coding, modulation, etc. However, most prior researches on blind recognition only focus on the scenario where the single signal processing method is applied, and it is impractical. In this paper, a blind recognition algorithm based on the characteristics of signal matrix and layered neural network for scrambled channel encoder is proposed. First, this method analyzes the difference between the matrix features in different types of scrambled channel encoders and regards it as an important condition for recognition. Then, the convolutional neural network is used to learn the scrambled codes and classify the different encoder types. At last, the layered neural network architecture of channel encoder type recognition is proposed, which can recognize different encoder types by categories. Moreover, it also can resist error code. In this paper, the recognition of 6 channel encoder types is realized with high robustness and adaptability. Simulation results show that the recognition rate of the convolution encoder has reached 95% when the bit error rate is 0.5%, which is better than the traditional channel encoder classification method based on matrix features.