基于网络度量的三分支孪生网络调制识别算法

doi:10.3778/j.issn.1002-8331.2006-0111

摘要/Abstract

摘要：

针对基于孪生网络的小样本调制识别算法存在相似识别类别混淆的问题，提出一种基于网络度量的三分支孪生网络调制识别算法。通过三分支孪生神经网络（Triplet Siamese Neural Network，TSN）将输入三元样本组的原始特征映射至特征子空间中。将正负样本与参考样本特征并联输入至两个参数共享的关系网络（Relation Network，RN），学习一个非线性的度量函数。通过各个类别的特征向量生成各类别的类原型，作为测试过程中的类特征输入。为降低信道噪声和信号接收误差对均值类原型表达的影响，采用局部异常因子算法（Local Outlier Factor，LOF）剔除类别中偏差数据。在公开的调制数据集DeepSig中进行验证，仿真结果表明，TSN-RN-LOF算法模型可以充分利用相似类别之间的差异信息，提取更具辨识度的特征，取得更优的识别性能。

关键词: 调制识别, 三分支孪生神经网络, 关系网络, 局部异常因子算法, 小样本

Abstract:

Aiming at the problem of similar recognition category confusion in small sample modulation recognition algorithms based on Siamese Neural Network, and the Triplet Siamese Network modulation recognition algorithm based on network measurement is proposed. Firstly, the original features of the input triplet set are mapped into the feature subspace through the Triplet Siamese Neural Network（TSN）. Then, the features of the positive and negative samples and the reference samples are input in parallel to the Relation Network（RN） shared by two parameters, and a nonlinear measure function is learned. Finally, the prototype of each category is generated through the feature vector of each category, which is used as class feature input in the testing process. At the time, in order to reduce the influence of channel noise and signal reception error on the expression of mean-type prototypes, the Local Outlier Factor（LOF） algorithm is used to eliminate the deviation data in the category. The results conducted in the public modulation data set DeepSig indicate that the algorithm model of TSN-RN-LOF makes full use of the difference information between similar categories, extracts more recognizable features and achieves better recognition performance.

Key words: modulation recognition, Triplet Siamese Neural Network（TSN）, Relation Network（RN）, Local Outlier Factor（LOF）, small sample

冯磊，蒋磊，许华，苟泽中. 基于网络度量的三分支孪生网络调制识别算法[J]. 计算机工程与应用, 2021, 57(19): 135-141.

FENG Lei, JIANG Lei, XU Hua, GOU Zezhong. Triplet Siamese Network Modulation Recognition Algorithm Based on Network Measurement[J]. Computer Engineering and Applications, 2021, 57(19): 135-141.

参考文献

[1] GIRSHICK R.Fast R-CNN[C]//2015 IEEE International Conference on Computer Vision，Santago，2015：1440-1448.
[2] HUANG G，LIU Z，LAURENS V D M，et al.Densely connected convolutional networks[C]//2017 IEEE Conference on Computer Vision and Pattern Recognition，Honolulu，2017：2261-2269.
[3] XIA Y，HE D，QIN T，et al.Dual learning for machine translation[C]//Proceedings of the the 30th Conference on Neural Information Processing Systems，Barcelona，2016：820-828.
[4] O’SHEA T J，CORGAN J，CLANCY T C.Convolutional radio modulation recognition networks[C]//Proceedings of the International Conference on Engineering Applications of Neural Networks，2016：213-226.
[5] 白芃远，许华，孙莉.基于卷积神经网络与时频图纹理信息的信号调制方式分类方法[J].西北工业大学学报，2019，37（4）：816-823.
BAI P Y，XU H，SUN L.A recognition algorithm for modulation schemes by convolution neural network and spectrum texture[J].Journal of Northwesteern Polytechnical University，2019，37（4）：816-823.
[6] SAK H，SENIOR A，BEAUFAYS F.Long short-term memory recurrent neural network architectures for large scale acoustic modeling[C]//Proceedings of the Fifteenth Annual Conference of The International Speech Communication Association，2014.
[7] 苟泽中，许华，郑万泽，等.基于半监督联合神经网络的调制识别算法[J].信号处理，2020，36（2）：168-176.
GOU Z Z，XU H，ZHENG W J，et al.Semi-supervised joint neural network based recognition algorithm of modulation signal[J].Journal of Signal Processing，2020，36（2）：168-176.
[8] DIXIT M，KWITT R，NIETHAMMER M，et al.AGA：Attribute-guided augmentation[C]//2017 IEEE Conference on Computer Vision and Pattern Recognition，2017：7455-7463.
[9] HARIHARAN B，GIRSHICK R.Low-shot visual recognition by shrinking and hallucinating features[C]//2017 IEEE International Conference on Computer Vision，2017：3037-3046.
[10] FINN C，ABBEEL P，LEVINE S.Model-agnostic meta-learning for fast adaptation of deep networks[C]//Proceedings of the 34th International Conference on Machine Learning，2017：1126-1135.
[11] SANTORO A，BARTUNOV S，BOTVINICK M，et al.Meta-learning with memory-augmented neural networks[C]//Proceedings of the 33rd International Conference on Machine Learning，2016：1842-1850.
[12] RAVI S，LAROCHELLE H.Optimization as a model for few-shot learning[C]//Proceedings of the International Conference on Learning Representations，Toulon，2017.
[13] SNELL J，SWERSKY K，ZEMEL R.Prototypical networks for few-shot learning[C]//Proceedings of the 31st Conference on Neural Information Processing Systems，2017：4080-4090.
[14] KOCH G，ZEMEL R，SALAKHUTDINOV R.Siamese neural networks for one-shot image recognition[C]//Proceedings of the ICML Deep Learning Workshop，2015.
[15] SUNG F，YANG Y，ZHANG L，et al.Learning to Compare：Relation network for few-shot learning[C]//Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognification，2017.
[16] SANDER J.LOF：Identifying density-based local outliers[J].Acmsigmod Record，2000，29（2）：93-104.
[17] O’SHEA T J，WEST N.Radio machine learning dataset generation with gnu radio[C]//Proceeding of the GNU Radio Conference，2016.
[18] HE K，ZHANG X，REN S，et al.Deep residual learning for image recognition[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition，2016：770-778.
[19] HERRMANS A，BEYER L，LEIBBE B.In defense of the triplet loss for person re-identifiction[J].arXiv：1703. 07737，2017.
[20] BISHOP C M.Pattern recognition and machine learning[M]//Information science and statistics.New York：Springer-Verlag，2006.