基于多模态特征融合网络的空时分组码识别算法

doi:10.12236/DZXB.20210326

摘要/Abstract

摘要：

针对现有算法在空时分组码（Space-Time Block Code， STBC）识别过程中，存在的低信噪比下误判概率高、识别效率低等问题，本文提出了一种基于多模态特征融合网络（Multi-Modality Features Fusion Network，MMFFN）的空时分组码自动识别方法.首先在合并卷积层将STBC时域样本映射为一维特征向量的基础上，采用多扩张率下的扩张卷积提取非连续时间窗的STBC码内特征，实现多时延特征自提取，然后构建多时序特征自提取模块以提取码间时序特征，进一步扩展映射特征类型，最后提取多时延拼接层的最大时延特征作为深层融合特征，并增加了带跨越连接的残差层以提升融合特征利用率，实现空时分组码识别.仿真实验结果表明，本文算法在-9dB下对6类STBC信号的识别准确率达到了90%以上，较现有识别算法的性能获得了显著提升，对低信噪比有较强的适应性.本文提出的STBC多时延特征提取和融合方法，为结合传统算法设计深度学习网络结构提供了新思路，其思想同样可应用于其他通信信号识别领域.

关键词: 空时分组码, 深度学习, 扩张卷积, 多时延特征, 多时序特征, 最大时延融合

Abstract:

* Aiming at the problems of the existing algorithms in the process of space-time block code (STBC) recognition, such as high misdiagnosis probability and low recognition efficiency under low signal to noise ratio (SNR), this paper proposes an automatic space-time block code recognition method based on multi-modality feature fusion network (MMFFN). Firstly, on the basis of mapping STBC time-domain samples into one-dimensional feature vectors by merging convolution layers, the dilated convolution at multiple dilation rates is used to extract STBC code features from discontinuous time windows, and the self-extraction of multi-delay features is realized. Then, the multi-sequence feature self-extraction module is constructed to extract the inter-code sequence feature, and the mapping feature types are further extended. Finally, the maximum delay feature of the multi-delay Mosaic layer is extracted as the deep fusion feature, and the residual layer with span connection is added to improve the utilization of fusion feature and realize space-time block code recognition. Simulation results show that the recognition accuracy of the proposed algorithm for 6 types of STBC signals reaches more than 90% under -9dB, which is significantly improved compared with the performance of existing recognition algorithms, and has a strong adaptability to low SNR. The STBC multi-delay feature extraction and fusion method proposed in this paper provides a new idea for the design of deep learning network structure by combining traditional algorithms, and the idea can also be applied to other communication signal recognition fields.

Key words: space-time block code, deep learning, dilated convolution, muti-delay features, muti-sequential features, maximum delay fusion

中图分类号:

TN911.7

张聿远, 闫文君, 张立民. 基于多模态特征融合网络的空时分组码识别算法[J]. 电子学报, DOI: 10.12236/DZXB.20210326.

ZHANG Yu-yuan, YAN Wen-jun, ZHANG Li-min. Space-Time Block Code Recognition Algorithm Based on Multi-Modality Features Fusion Network[J]. Acta Electronica Sinica, DOI: 10.12236/DZXB.20210326.

图/表 11

图1 多模态特征融合网络框架

图2 空时分组码相关性分布

表1 STBC多时延相关性分布

STBC类型	时延 $τ = 1$	时延 $τ = 2$	时延 $τ = 4$
SM	×	×	×
AL	√	×	×
STBC3-1	√	√	×
STBC3-2	√	√	×
STBC3-3	√	√	√
STBC4	√	√	√

表1 STBC多时延相关性分布

STBC类型	时延 $τ = 1$	时延 $τ = 2$	时延 $τ = 4$
SM	×	×	×
AL	√	×	×
STBC3-1	√	√	×
STBC3-2	√	√	×
STBC3-3	√	√	√
STBC4	√	√	√

表2 MMFFN网络参数设置

模块名称	参数设置
多时延特征自提取	输入层:input = $2 × 128$
	合并卷积层:filters = 128, kernel size = (2,1), padding = 'valid'
	扩张卷积层:filters = 64, kernel size = (1,2), dilation_rate = (1,1), padding = 'same'	扩张卷积层:filters = 64, kernel size = (1,2), dilation_rate = (1,2), padding = 'same'	扩张卷积层:filters = 64, kernel size = (1,2), dilation_rate = (1,4), padding = 'same'
多时序特征自提取	卷积层:filters = 128, kernel size = (1,4), padding = 'same'	卷积层:filters = 128, kernel size = (1,4), padding = 'same'	卷积层:filters = 128, kernel size = (1,4), padding = 'same'
	尺度变换:(128, 1, 128)→(128, 128)	尺度变换:(128, 1, 128)→(128, 128)	尺度变换:(128, 1, 128)→(128, 128)
	LSTM层:units = 32, recurrent_dropout=0.5, return_sequences=True	LSTM层:units = 32, recurrent_dropout=0.5, return_sequences=True	LSTM层:units = 32, recurrent_dropout=0.5, return_sequences=True
	尺度变换:(128, 32)→(32, 1,128) (X_0)	尺度变换:(128, 32)→(32, 1,128) (X_1)	尺度变换:(128, 32)→(32, 1,128) (X_2)
最大时延特征融合模块	拼接层:concatenate ([X_0, X_1, X_2], axis=2) (X_shortcut)
	最大时延融合层:MaxPooling2D (pool_size = (3, 1), padding = 'valid')
	卷积层:filters = 32, kernel size = (1,2), padding = 'same'
	卷积层:filters = 32, kernel size = (1,2), padding = 'same' (X_3)
	跨越连接层:layers.add ([X_3, X_ shortcut])
	全连接层:units = 64, ReLU激活函数
	全连接层:units = 6, Softmax激活函数

表2 MMFFN网络参数设置

模块名称	参数设置
多时延特征自提取	输入层:input = $2 × 128$
	合并卷积层:filters = 128, kernel size = (2,1), padding = 'valid'
	扩张卷积层:filters = 64, kernel size = (1,2), dilation_rate = (1,1), padding = 'same'	扩张卷积层:filters = 64, kernel size = (1,2), dilation_rate = (1,2), padding = 'same'	扩张卷积层:filters = 64, kernel size = (1,2), dilation_rate = (1,4), padding = 'same'
多时序特征自提取	卷积层:filters = 128, kernel size = (1,4), padding = 'same'	卷积层:filters = 128, kernel size = (1,4), padding = 'same'	卷积层:filters = 128, kernel size = (1,4), padding = 'same'
	尺度变换:(128, 1, 128)→(128, 128)	尺度变换:(128, 1, 128)→(128, 128)	尺度变换:(128, 1, 128)→(128, 128)
	LSTM层:units = 32, recurrent_dropout=0.5, return_sequences=True	LSTM层:units = 32, recurrent_dropout=0.5, return_sequences=True	LSTM层:units = 32, recurrent_dropout=0.5, return_sequences=True
	尺度变换:(128, 32)→(32, 1,128) (X_0)	尺度变换:(128, 32)→(32, 1,128) (X_1)	尺度变换:(128, 32)→(32, 1,128) (X_2)
最大时延特征融合模块	拼接层:concatenate ([X_0, X_1, X_2], axis=2) (X_shortcut)
	最大时延融合层:MaxPooling2D (pool_size = (3, 1), padding = 'valid')
	卷积层:filters = 32, kernel size = (1,2), padding = 'same'
	卷积层:filters = 32, kernel size = (1,2), padding = 'same' (X_3)
	跨越连接层:layers.add ([X_3, X_ shortcut])
	全连接层:units = 64, ReLU激活函数
	全连接层:units = 6, Softmax激活函数

图3 不同时延特征网络的识别性能

图4 不同时延特征网络的混淆矩阵

图5 不同时序与时延特征网络的识别性能

图6 LSTM层结构参数对识别性能的影响

图7 不同识别算法的性能对比

表3 网络复杂度分析

网络	空间复杂度	时间复杂度(ms)
MMFFN	476,998	0.106 ms
FDSCF+CNN	2,645,078	0.015 ms
CNN+LSTM	1,059,522	0.319 ms

图8 不同样本数下的识别性能对比

参考文献 18

1	ELDEMERDASH Y A, DOBRE O A, ÖNER M. Signal identification for multiple-antenna wireless systems: achievements and challenges[J]. IEEE Communications Surveys & Tutorials, 2016, 18(3): 1524-1551.
2	KARAMI E, DOBRE O A. Identification of SM-OFDM and AL-OFDM signals based on their second-order cyclostationarity [J]. IEEE Transactions on Vehicular Technology, 2015, 64(3): 942-953.
3	LING Q, ZHANG L, YAN W, et al. Hierarchical space-time block codes signals classification using higher order cumulants [J]. Chinese Journal of Aeronautics, 2016, 29(3): 754-762.
4	闫文君, 张立民, 凌青, 等. 基于高阶统计特征的空时分组码盲识别方法[J]. 电子与信息学报, 2016, 38(3): 668-673.
	YAN W J, ZHANG L M, LING Q, et al. An algorithm for blind classification of space-time block code based on higher-order statistics [J]. Journal of Electronics & Information Technology, 2016, 38(3): 668-673. (in Chinese)
5	DEHRI B, BESSEGHIER M, DJEBBAR A B, et al. Blind digital modulation classification for STBC-OFDM system in presence of CFO and channels estimation errors [J]. IET Communications, 2019, 13(17): 2827-2833.
6	KHOSRAVIYANI M, KALBKHANI H, SHAYESTEH M G. Higher order statistics for modulation and STBC recognition in MIMO systems [J]. IET Communications, 2019, 13(16): 2436-2446.
7	张天琪, 范聪聪, 喻盛琪, 等. 基于JADE与特征提取的正交/非正交空时分组码盲识别[J]. 系统工程与电子技术, 2020, 42(4): 933-939.
	ZHANG T Q, FAN C C, YU S Q, et al. Blind recognition of orthogonal and non-orthogonal space-time block codes based on JADE and feature extraction[J]. Systems Engineering and Electronics, 2020, 42(4): 933-939. (in Chinese)
8	YAN W, LING Q, ZHANG L, et al. Convolutional neural networks for space-time block coding recognition [EB/OL].(2019-10-09)[2021-03-09]..
9	于柯远, 张立民, 闫文君, 等.基于深度学习的多STBC盲识别算法[J/OL].系统工程与电子技术. .
	YU K Y, ZHANG L M, YAN W J, et al. Blind recognition algorithm for multi-STBC based on deep learning [J/OL]. Systems Engineering and Electronics. (in Chinese)
10	张聿远, 闫文君, 林冲, 等. 利用卷积-循环神经网络的串行序列空时分组码识别方法[J]. 信号处理, 2021, 37(1): 19-27.
	ZHANG Y Y, YAN W J, LIN C, et al. Serial sequence space-time block code recognition method by using convolutional- recurrent neural networks [J]. Journal of Signal Processing, 2021, 37(1): 19-27. (in Chinese)
11	O'SHEA T J, CORGAN J, CLANCY T C. Convolutional radio modulation recognition networks [C]//International Conference on Engineering Applications of Neural Networks. Aberdeen: Springer; 2016: 213-226.
12	ZHANG Z F, WANG C, GAN C Q, et al. Automatic modulation classification using convolutional neural network with features fusion of SPWVD and BJD [J]. IEEE Transactions on Signal and Information Processing over Networks, 2019, 5(3): 469-478.
13	ZHENG S L, QI P H, CHEN S C, et al. Fusion methods for CNN-based automatic modulation classification[J]. IEEE Access, 2019, 7: 66496-66504.
14	BU K, HE Y, JING X J, et al. Adversarial transfer learning for deep learning based automatic modulation classification[J]. IEEE Signal Processing Letters, 2020, 27: 880-884.
15	TUNZE G B, HUYNH-THE T, LEE J M, et al. Sparsely connected CNN for efficient automatic modulation recognition [J]. IEEE Transactions on Vehicular Technology, 2020, 69(12): 15557-15568.
16	ZHAO S Q, WANG W H, ZENG D G, et al. A novel aggregated multi-path extreme gradient boosting approach for radar emitter classification[J/OL]. IEEE Transactions on Industrial Electronics, 2021, doi: 10.1109/TIE.2021.3055155 .
17	王传旭, 薛豪. 基于GFU和分层LSTM的组群行为识别研究方法[J]. 电子学报, 2020, 48(8): 1465-1471.
	WANG C X, XUE H. Group activity recognition based on GFU and hierarchical LSTM[J]. Acta Electronica Sinica, 2020, 48(8): 1465-1471. (in Chinese)
18	王传旭, 胡小悦, 孟唯佳, 等. 基于多流架构与长短时记忆网络的组群行为识别方法研究[J]. 电子学报, 2020, 48(4): 800-807.
	WANG C X, HU X Y, MENG W J, et al. Research on group behavior recognition method based on multi-stream architecture and long short-term memory network[J]. Acta Electronica Sinica, 2020, 48(4): 800-807. (in Chinese)

[1]	吴靖, 叶晓晶, 黄峰, 陈丽琼, 王志锋, 刘文犀. 基于深度学习的单帧图像超分辨率重建综述[J]. 电子学报, 2022, 50(9): 2265-2294.
[2]	李雪莹, 王田路, 梁鹏, 王翀. 基于系统模型的用户评论中非功能需求的自动分类[J]. 电子学报, 2022, 50(9): 2079-2089.
[3]	琚长瑞, 秦晓燕, 袁广林, 李豪, 朱虹. 尺度敏感损失与特征融合的快速小目标检测方法[J]. 电子学报, 2022, 50(9): 2119-2126.
[4]	张志昌, 于沛霖, 庞雅丽, 朱林, 曾扬扬. SMGN：用于对话状态跟踪的状态记忆图网络[J]. 电子学报, 2022, 50(8): 1851-1858.
[5]	张亚洲, 俞洋, 朱少林, 陈锐, 戎璐, 梁辉. 一种量子概率启发的对话讽刺识别网络模型[J]. 电子学报, 2022, 50(8): 1885-1893.
[6]	王飞扬, 冀鹏欣, 孙笠, 危倩, 李根, 张忠宝. 一种基于深度学习的动态社交网络用户对齐方法[J]. 电子学报, 2022, 50(8): 1925-1936.
[7]	徐兴荣, 刘聪, 李婷, 郭娜, 任崇广, 曾庆田. 基于双向准循环神经网络和注意力机制的业务流程剩余时间预测方法[J]. 电子学报, 2022, 50(8): 1975-1984.
[8]	欧阳与点, 谢鲲, 谢高岗, 文吉刚. 面向大规模网络测量的数据恢复算法：基于关联学习的张量填充[J]. 电子学报, 2022, 50(7): 1653-1663.
[9]	裴炤, 邱文涛, 王淼, 马苗, 张艳宁. 基于Transformer动态场景信息生成对抗网络的行人轨迹预测方法[J]. 电子学报, 2022, 50(7): 1537-1547.
[10]	彭闯, 王伦文, 胡炜林. 融合深度特征的电磁频谱异常检测算法[J]. 电子学报, 2022, 50(6): 1359-1369.
[11]	杨伟超, 杜宇, 文伟, 侯舒维, 徐常志, 张建华. 基于多重分形谱智能分析的卫星信号调制识别研究[J]. 电子学报, 2022, 50(6): 1336-1343.
[12]	李政伟, 李佳树, 尤著宏, 聂茹, 赵欢, 钟堂波. 基于异质图注意力网络的miRNA与疾病关联预测算法[J]. 电子学报, 2022, 50(6): 1428-1435.
[13]	张波, 陆云杰, 秦东明, 邹国建. 一种卷积自编码深度学习的空气污染多站点联合预测模型[J]. 电子学报, 2022, 50(6): 1410-1427.
[14]	廖勇, 李玉杰. 一种轻量化低复杂度的FDD大规模MIMO系统CSI反馈方法[J]. 电子学报, 2022, 50(5): 1211-1217.
[15]	冀振燕, 韩梦豪, 宋晓军, 冯其波. 面向激光光条图像修复的循环相似度映射网络[J]. 电子学报, 2022, 50(5): 1234-1242.