Deep Gated Recurrent Unit Convolution Network for Radio Signal Recognition

This paper proposes a Deep Gated Recurrent Unit (GRU) Convolution Network for feature extraction and classification of radio signals. First, a set of 1-D filters are constructed to extract the hierarchical features of raw signals. Then a GRU layer is designed to explore the low-term memory for dealing with sequence signals. This network can not only extract discriminant features of raw signals, but also deal with sequence signals. The authors evaluate the proposed method on 31 categories of signal which have different modulation and channel coding types, and the results show that the significant performance gains can be obtained.