28 September 2023 in IEEE Transactions on Cognitive Communications and Networking
In modern communication systems, automatic modulation recognition (AMR) has broad application prospects but a large amount of training data is usually required. Due to the high cost of adding annotations to signals, it is of great significance to study AMR under the condition of few labeled samples. In this paper, an end-to-end semi-supervised radio signal recognition algorithm is proposed. In view of the poor anti-noise performance of the original IQ data, short-time Fourier transform (STFT) is introduced to mine the stable modulation characteristics. In order to overcome the noise problem caused by pseudo-labels, a mixed sample data augmentation (MSDA) method PFMix is proposed, and the consistency constraint of the model disturbance based on mean teacher (MT) framework and data disturbance based on PFMix is designed. Meanwhile, in order to provide pseudo-labels with high confidence, entropy regularization is constructed by taking the prediction variance of MT framework as correction factor. The experimental results using the open source RML2016.10A dataset show the superior performance of the proposed method. The average recognition accuracy with 1% labeled data is 88.82% at 0 dB, which is a significant improvement compared with state of the art in semi-supervised AMR.
19 September 2022 in IEEE Transactions on Instrumentation and Measurement
Radar signal recognition is of great significance for threat analysis and intelligence support in radar electronic warfare. However it remains a challenging task due to (1) the variety of radar signal modulation types, (2)the increasingly saturated electromagnetic environment and the more serious influence of noise interference, and (3)the intrinsic real-time requirement of this application. In this paper, a novel automatic modulation recognition algorithm based on the joint architecture of convolution neural network (CNN) and the support vector machine (SVM) is proposed to comprehensively address these challenges. In particular, this algorithm first employs time-frequency analysis to better show the pulse representation in the time-frequency domain. Then a well-designed DCNN equipped with a hybrid attention (HA) mechanism and skip feature aggregation (SFA) is used to automatically learn the feature representations of distinctiveness from time frequency images. HA is designed to quickly capture the important information of local parts in sight and enhance the ability of the backbone network to mine high-level features. SFA is applied to selectively aggregate features among channels for self-monitoring, which improves the expression and generalization ability. Sixteen types of radar signals are used to verify the feasibility and effectiveness of this algorithm. The experimental evaluation shows that the proposed method not only achieve competitive accuracy with a better timeliness than state-of-the-art methods, but also achieves effective identification of most of the signal modulation types.