Modulation Recognition of Communication Signals Based on Multimodal Feature Fusion

Abstract

Modulation recognition is the indispensable part of signal interception analysis, which has always been the research hotspot in the field of radio communication. With the increasing complexity of the electromagnetic spectrum environment, interference in signal propagation becomes more and more serious. This paper proposes a modulation recognition scheme based on multimodal feature fusion, which attempts to improve the performance of modulation recognition under different channels. Firstly, different time- and frequency-domain features are extracted as the network input in the signal preprocessing stage. The residual shrinkage building unit with channel-wise thresholds (RSBU-CW) was used to construct deep convolutional neural networks to extract spatial features, which interact with time features extracted by LSTM in pairs to increase the diversity of the features. Finally, the PNN model was adapted to make the features extracted from the network cross-fused to enhance the complementarity between features. The simulation results indicated that the proposed scheme has better recognition performance than the existing feature fusion schemes, and it can also achieve good recognition performance in multipath fading channels. The test results of the public dataset, RadioML2018.01A, showed that recognition accuracy exceeds 95% when the signal-to-noise ratio (SNR) reaches 8 dB.

Universal and complementary representation learning for automatic modulation recognition

“…For fusion DL-AMR methods, we analyze and compare them in three categories in Figure 4b. I), Direct concentration or fixed-order fusion such as WSMF [12] and optimized PNN model [9]. However, without further capturing the underlying information, these methods are at least 2% lower than the UCNet.…”

“…Modality embedding: Inspired by [9], the original signal is transferred into three modalities, i.e. IQ, AP, and SP.…”

IDAF: Iterative Dual-Scale Attentional Fusion Network for Automatic Modulation Recognition

“…A/P Long Short Term Memory (LSTM) [16], a LSTM denoising auto-encoder [14] Well recognize AM-SSB, and distinguish between QAM16 and QAM64 [22] Spectrum RSBU-CW with Welch spectrum, square spectrum, and fourth power spectrum [23]; SCNN [18] with the short-time Fourier transform (STFT), a fine-tuned CNN model [17] with smooth pseudo-Wigner-Ville distribution and Born-Jordan distribution Achieves high accuracy of PSK [23], recognizes OFDM well, which is revealed only in the spectrum domain due to its plentiful sub-carriers [17] In recent years, several studies have also focused on the advantages of multimodal information fusion for AMR tasks. In [24], modality discriminative features are captured separately using three Resnet networks, and I/Q, A/P, and the amplitudes of spectrum, square spectrum, and fourth power spectrum features are concatenated with the corresponding bitwise summation.…”

“…Unlike the above direct addition or multiplication fusion approach, Ref. [23] uses a PNN (Product-based Neural Network) model to cross-fuse the three modal features in a fixed order. However, most of the above methods fuse multimodal features via direct or crosswise summation or outer product, which tends to ignore the variability of different modes and their different impacts on modulation identification.…”

Universal and Complementary Representation Learning for Automatic Modulation Recognition

“…Qi et al propose a Waveform Spectrum Multi-modality Fusion (WSMF) method, which relies on a deep Residual Network (ResNet) and a concise concatenation layer [8]. An optimized Product-based Neural Networks (PNN) model [9] cross-combines the features extracted from I/Q, A/P, and spectrum. However, the above methods simply carry out the crossconnect or direct concatenation of features instead of further capturing the underlying information.…”