Signal Modulation Recognition Algorithm Based on Improved Spatiotemporal Multi-Channel Network

• DOI: 10.3390/electronics12020422
• Publication Date: 2023-01-16T09:06:47Z
• Authors: Shunhu Hou; Youchen Fan; Bing Han; Yuhai Li; Shengliang Fang
• Keywords: Overfitting; Robustness; Kernel (algebra); Feature (linguistics); Convolution (computer science); Modulation (music)

Automatic modulation recognition (AMR) plays an essential role in modern communication systems. In recent years, various algorithms based on deep learning have been emerging, but the problem of low accuracy has not solved well. To solve this problem, existing MCLDNN algorithm, paper, we proposed improved spatiotemporal multi-channel network (IQ-related features Multi-channel Convolutional Bi-LSTM with Gaussian noise, IQGMCL). Firstly, dividing input IQ signals into three channels, time sequence feature extraction is carried out for route I, Q, and IQ, respectively. For convolution kernel (2,1) first used to extract relevant features. Two layers small (1,3) are further, channels further. Then, a two-layer short-length memory from space more effectively. Through comparison experiments, introduced replace one layer LSTM, fully connected removed prevent overfitting. Finally, multiplicative noise naturally corrode parameters, further improving robustness model. Experiments public datasets RML2016.10a, RML2016.10b, RML2016.04C. The experiments show that IQGMCL higher accuracies all datasets, especially RML2016.10a dataset. When SNR 4 dB, reaches 93.52%. greater than 0 average 92.3%, 1.31%, 1.2% original network,