基于元学习的时变信道估计方法

doi:10.12305/j.issn.1001-506X.2023.06.32

摘要/Abstract

摘要：

为了克服现有的基于深度学习的信道估计方法存在的训练样本与时间开销过大、且离线训练的网络无法适应在实际情况下实时变化的信道环境的问题, 提出了一种高速移动正交频分复用(orthogonal frequency division multiplexing, OFDM)系统中的基于元学习的时变信道估计方法。该方法利用模型无关元学习(model-agnostic meta-learning, MAML)方法对不同的信道任务进行线下训练, 仅利用少量的训练样本即可使网络充分学习到信道的传输特性，以及具备快速适应新任务的能力, 且具有较低的计算复杂度。在线下训练中, 该方法将网络的训练目标设置为具有较高精度的信道估计, 而非理想的信道信息, 增强了估计模型的实用性。另外, 该方法仅采用接收的导频信号进行线下训练与线上估计, 减少了网络输入样本的数目, 进一步降低了计算复杂度。仿真结果表明, 所提方法具有较高的估计精度与较低的计算复杂度, 且可以快速地适应新的信道环境, 适用于在高速移动通信系统中获取时变信道。

关键词: 高速移动, 正交频分复用, 时变信道估计, 元学习

Abstract:

In the existing deep-learning-based channel estimation methods, there are problems such as large training data and huge time overhead, and the offline training network cannot adapt to the actual real-time changing channel environment. To overcome the problems above, a meta-learning-based time-varying channel estimation method in high-speed mobile orthogonal frequency division multiplexing (OFDM) system is proposed in this paper. The model-agnostic meta-learning (MAML) method is used for offline training for different sub-tasks, so that the network can adequately learn the characteristics of channel transmission and have the ability to quickly adapt to new tasks only by few training samples, which makes it have low computational complexity. In offline training, the proposed method sets the training target of the network as channel estimation with high accuracy rather than ideal channel information, which enhances the practicability of the estimation model. In addition, only the received pilot signal is used for offline training and online estimation, which reduces the number of network input samples and further reduces the computational complexity. Simulation results show that the proposed method has high estimation accuracy and low computational complexity, and it can quickly adapt to the new channel environment, which is suitable for time-varying channels acquisition in high-speed mobile communication systems.

Key words: high-speed mobile, orthogonal frequency division multiplexing(OFDM), time-varying channel estimation, meta-learning

中图分类号:

TN929.5

杨丽花, 任露露, 呼博, 邵永琪, 聂倩. 基于元学习的时变信道估计方法[J]. 系统工程与电子技术, 2023, 45(6): 1872-1879.

Lihua YANG, Lulu REN, Bo HU, Yongqi SHAO, Qian NIE. Meta-learning based time-varying channel estimation method[J]. Systems Engineering and Electronics, 2023, 45(6): 1872-1879.

图/表 10

图1

表1

表2

图2

图3

图4

图5

图6

图7

表3

参考文献 31

1	MEHRABI M, MOHAMMADKARIMI M, ARDAKANI M, et al. A deep learning-based channel estimation for high mobility vehicular communications[C]//Proc. of the IEEE International Conference on Computing, Networking and Communications, 2020: 2325-2626.
2	HU Y , LI H Y , CHANG Z , et al. Scheduling strategy for multimedia heterogeneous high-speed train networks[J]. IEEE Trans.on Vehicular Technology, 2017, 66 (4): 3265- 3279. doi: 10.1109/TVT.2016.2587080
3	WU J X , FAN P E . A survey on high mobility wireless communications: Challenges, opportunities and solutions[J]. IEEE Access, 2017, 4, 450- 476.
4	RAMTEKE S B, DESHMUKH A Y, DEKATE K N. A review on design and analysis of 5G mobile communication mimo system with ofdm[C]//Proc. of the 2nd International Conference on Electronics, Communication and Aerospace Technology, 2018: 542-546.
5	LU L , LI G Y , SWINDLEHURST A L , et al. An overview of massive mimo: Benefits and challenges[J]. IEEE Journal of Selected Topics in Signal Processing, 2014, 8 (5): 742- 758. doi: 10.1109/JSTSP.2014.2317671
6	CHENG W C , ZHANG X , ZHANG H L . Statistical-QoS driven energy-efficiency optimization over green 5G mobile wireless networks[J]. IEEE Journal on Selected Areas in Communications, 2016, 34 (12): 3092- 3107.
7	GHETH W , RABIE K M , ADEBISI B , et al. Communication systems of high-speed railway: a survey[J]. Trans.on Emerging Telecommunications Technologies, 2021, 32 (4): e4189.
8	CHEN J Q , GE X H , NI Q , et al. Coverage and handoff analysis of 5G fractal small cell networks[J]. IEEE Trans.on Wireless Communications, 2019, 18 (2): 1263- 1276. doi: 10.1109/TWC.2018.2890662
9	XIN W, TAN Z H, XIA C. Doppler diversity receiver for broadband wireless OFDM system under high-speed mobile environments[C]//Proc. of the International Symposium on Microwave, 2006: 1444-1447.
10	HLAWATSCH F , MATZ G . Wireless communications over rapidly time-varying channels[M]. New York: Academic Press, 2011.
11	GE X H , QIU Y H , CHEN J Q , et al. Wireless fractal cellular networks[J]. IEEE Wireless Communications, 2016, 23 (5): 110- 119. doi: 10.1109/MWC.2016.7721749
12	SENOL H , TEPEDELENLIOGLU C . Subspace-based estimation of rapidly varying mobile channels for OFDM systems[J]. IEEE Trans.on Signal Processing, 2020, 69, 385- 400.
13	LYU X T , FENG W , GE N , et al. Deep learning-based symbol detection for time-varying nonstationary channels[J]. China Communications, 2022, 19 (3): 14- 15.
14	CHAI L F, ZHANG Y F, ZHANG Z Z. The research on acoustic communication anti-Doppler shift technology based on FrFFT[C]//Proc. of the IEEE International Conference on Signal, Information and Data Processing, 2019.
15	HRYCAK T , DAS S , MATZ G , et al. Practical estimation of rapidly varying channels for ofdmsystems[J]. IEEE Trans.on Communications, 2011, 59 (11): 3040- 3048. doi: 10.1109/TCOMM.2011.082111.110075
16	YE H , LI G Y . Power of deep learning for channel estimation and signal detection in ofdmsystems[J]. IEEE Wireless Communication Letters, 2017, 7 (1): 114- 117.
17	MA X L, HAO Y, YE L. Learning assisted estimation for time-varying channels[C]//Proc. of the 15th International Symposium on Wireless Communication Systems, 2018.
18	YANG Y W , GAO F F , MA X L , et al. Deep learning-based channel estimation for doubly selective fading channels[J]. IEEE Access, 2019, 7, 36579- 36589. doi: 10.1109/ACCESS.2019.2901066
19	LIAO Y, HUA Y X, DAI X W, et al. ChanEstNet: A deep learning-based channel estimation for high-speed scenarios[C]//Proc. of the IEEE International Conference on Communications, 2019.
20	SOLTANI M , POURAHMADJ M , MIRZAEI A , et al. Deep learning-based channel estimation[J]. IEEE Communications Letters, 2019, 23 (4): 652- 655. doi: 10.1109/LCOMM.2019.2898944
21	张庆林, 王永明. 元学习能力及其培养[M]. 北京: 中国教育学刊, 1996: 34- 37.
	ZHANG Q L , WANG Y M . Meta-learning ability and its cultivation[M]. Beijing: Chinese Journal of Education, 1996: 34- 37.
22	彭慧民. 元学习: 基础与应用[M]. 北京: 电子工业出版社, 2021.
	PENG H M . Meta-learning: Fundamentals and applications[M]. Beijing: Publishing House of Electronics Industry, 2021.
23	MAO H X, LU H C, LU Y J, et al. RoemNet: robust meta-learning-based channel estimation in ofdmsystems[C]//Proc. of the IEEE International Conference on Communications, 2019.
24	PARK S , JANG H , SIMEONE O , et al. Learning to demodulate from few pilots via offline and online meta-learning[J]. IEEE Trans.on Signal Processing, 2021, 69, 226- 239. doi: 10.1109/TSP.2020.3043879
25	MEI K , LIU J , ZHANAG X Y , et al. A low complexity learning-based channel estimation for ofdmsystems with online training[J]. IEEE Trans.on Communications, 2021, 69 (10): 6722- 6733. doi: 10.1109/TCOMM.2021.3095198
26	SETA K, IKEDA M. Model based development of a meta-learning support system to prompt self-awareness through presentation for meta-learning[C]//Proc. of the 3rd International Conference on Awareness Science and Technolgy, 2011.
27	HOSPEDALES T M , ANTONIOU A , MICAELLI P , et al. Meta-learning in neural networks: A survey[J]. IEEE Trans.on Pattern Analysis and Machine Intelligence, 2022, 44 (9): 5149- 5169.
28	PARK S, SIMEONE O, KANG J. End-to-end fast training of communication links without a channel model via online meta-learning[C]//Proc. of the IEEE 21st International Workshop on Signal Processing Advances in Wireless Communications, 2020.
29	RAHMAN M H, SHAHJALAL M, ALI M O, et al. Deep learning-based pilot assisted channel estimation for rician fading Massive mimouplink communication system[C]//Proc. of the 12th International Conference on Ubiquitous and Future Networks, 2021.
30	YIN M Z, TUCKER G, ZHOU M Y, et al. Meta-learning without memorization[C]//Proc. of the International Confe-rence on Learning Representations, 2020.
31	AOUDIA F A , HOYDIS J . Model-free training of end-to-end communication systems[J]. IEEE Journal on Selected Areas in Communications, 2019, 37 (11): 2503- 2516. doi: 10.1109/JSAC.2019.2933891

参数	取值
调制方式	QPSK与16QAM
子载波数	128
CP长度	16
导频数	32
均衡方法	迫零(ZF)
载波频率/GHz	2.35
载波间隔/kHz	15
归一化多普勒频移f_d	0.003∶0.003∶0.3
信道任务数U	100
基础学习器学习率α	0.001
元学习器学习率β	0.003
微调网络学习率γ	0.001

抽头数	时延/ns	功率/dB	多普勒谱
1	0	0	莱斯
2	710	-6.51	经典
3	1 090	-16.27	经典
4	1 730	-25.71	经典
5	2 910	-29.31	经典

估计方法		复杂度/FLOPs
估计方法		每秒浮点运算次数	计算复杂度
基于BP	线下训练	8G(NE)	2.621 44×10⁸
基于BP	线上估计	4NE	1.310 72×10⁵
LSTM	线下训练	2G(4(2NE+E²+2N)+4N²+2N)	1.182 208×10⁹
LSTM	线上估计	4(2NE+E²+2N)+4N²+2N	5.911 04×10⁵
新方法	线下训练	4G(N_pE)	3.276 8×10⁷
新方法	线上估计	4N_pE	3.276 8×10⁴

[1]	汪锐, 张天骐, 安泽亮, 王雪怡, 方竹. 基于联合特征参数和一维CNN的MIMO-OFDM系统调制识别算法[J]. 系统工程与电子技术, 2023, 45(3): 902-912.
[2]	吕岩, 曹菲, 许剑锋, 冯晓伟. 基于FRFT的单基地MIMO雷达稳健波束形成算法[J]. 系统工程与电子技术, 2023, 45(1): 79-85.
[3]	聂倩, 杨丽花, 呼博, 任露露. 基扩展模型下基于LSTM神经网络的时变信道预测方法[J]. 系统工程与电子技术, 2022, 44(9): 2971-2977.
[4]	郭军成, 万刚, 胡欣杰, 严发宝, 王帅. 太阳射电频谱爆发识别的元学习方法[J]. 系统工程与电子技术, 2022, 44(8): 2410-2418.
[5]	王震铎, 朱云飞, 宁晓燕, 刁鸣. FrFT-OFDM系统中联合ICI自消除方案[J]. 系统工程与电子技术, 2022, 44(8): 2645-2651.
[6]	孙晶明, 虞盛康, 孙俊. 基于元学习的雷达小样本目标识别方法及改进[J]. 系统工程与电子技术, 2022, 44(6): 1839-1845.
[7]	高潭, 吕成财, 田川. 面向OFDM-MFSK水声通信的差错控制方法[J]. 系统工程与电子技术, 2022, 44(5): 1701-1708.
[8]	张捷, 杨丽花, 聂倩. 新型的基于堆栈式ELM的时变信道预测方法[J]. 系统工程与电子技术, 2022, 44(2): 662-667.
[9]	刘步花, 丁丹, 杨柳, 薛乃阳, 刘仲谦. 基于DNN的无人机数据OFDM传输技术[J]. 系统工程与电子技术, 2022, 44(2): 696-702.
[10]	刘旗, 张新禹, 刘永祥. 基于门控多尺度匹配网络的小样本SAR目标识别[J]. 系统工程与电子技术, 2022, 44(11): 3346-3356.
[11]	朱立为, 黄知涛. 多载波OFDM信号识别方法[J]. 系统工程与电子技术, 2022, 44(11): 3522-3528.
[12]	王华华, 李延山, 余永坤. 低分辨率ADC下MIMO-OFDM系统中的广义Turbo信号检测[J]. 系统工程与电子技术, 2022, 44(1): 299-306.
[13]	季策, 王金芝, 李伯群. 基于RSAMP算法的OFDM稀疏信道估计[J]. 系统工程与电子技术, 2021, 43(8): 2290-2296.
[14]	庄陵, 叶华双. 改进的压缩感知限幅噪声消除方案[J]. 系统工程与电子技术, 2021, 43(8): 2341-2346.
[15]	李睿峰, 许爱强, 孙伟超, 王树友. 基于元学习的航空电子设备特征选择算法推荐方法[J]. 系统工程与电子技术, 2021, 43(7): 2011-2020.