Combinatorial neural network signal modulation recognition algorithm based on attention mechanism

Abstract

Aiming at the problems of low recognition rate and confused signal classification of deep learning modulation recognition network, combined neural network of one-dimensional residual network and long short-term memory based on efficient channel attention (ECA-RLDNet) is proposed. The algorithm designs a one-dimensional efficient channel attention mechanism to connect two feature extraction network units, uses the one-dimensional residual network to extract signal time series features, the attention mechanism gives higher weight to the key information of signal features, and further uses the long short-term memory to extract time series association features to obtain comprehensive and effective feature information. By simulating the modulation signal dataset under non-ideal channel and experimenting with the algorithm, the experimental results indicate that the highest recognition accuracy of ECA-RLDNet reaches 92.32%, which reduces the probability of confusion of high-order digital modulated signals.