基于双编码器<italic style="font-style: italic">β</italic>-VAE与注意力机制的单通道盲分离算法

高勇; 罗廷艺

doi:10.11959/j.issn.1000-436x.TXXB260086

您当前的位置：

首页 >

文章列表页 >

基于双编码器β-VAE与注意力机制的单通道盲分离算法

更新时间：2026-04-09

- 基于双编码器β-VAE与注意力机制的单通道盲分离算法
- Single-channel blind source separation algorithm based on a dual-encoder β-VAE and attention mechanisms
- 通信学报 2026年
- 作者机构：
  
  四川大学电子信息学院，四川成都 610065
- 作者简介：
  
  高勇，gaoyong@scu.edu.cn
- 基金信息：
  
  通信抗干扰全国重点实验室2024年度基础科研创新基金（稳定支持）项目(IFN202403)
- DOI：10.11959/j.issn.1000-436x.TXXB260086
  中图分类号： TN911.7
- 收稿：2026-02-06，
  
  修回：2026-04-09，
  
  录用：2026-04-09，
- 稿件说明：
移动端阅览
高勇, 罗廷艺. 基于双编码器β-VAE与注意力机制的单通道盲分离算法[J/OL]. 通信学报, 2026.

GAO Yong, LUO Tingyi. Single-channel blind source separation algorithm based on a dual-encoder β-VAE and attention mechanisms[J/OL]. Journal on Communications, 2026.
高勇, 罗廷艺. 基于双编码器β-VAE与注意力机制的单通道盲分离算法[J/OL]. 通信学报, 2026. DOI： 10.11959/j.issn.1000-436x.TXXB260086.

GAO Yong, LUO Tingyi. Single-channel blind source separation algorithm based on a dual-encoder β-VAE and attention mechanisms[J/OL]. Journal on Communications, 2026. DOI： 10.11959/j.issn.1000-436x.TXXB260086.

摘要

针对单通道盲源分离中传统方法误码率高、计算复杂度大等问题，提出基于双编码器

-VAE与注意力机制的单通道盲分离算法。该算法采用双编码器并行结构增强信号特征提取能力，引入膨胀卷积提取多尺度特征，利用Conformer模块融合局部与全局特征，在解码器中加入ECA模块以提升重构质量，并采用多损失联合优化模型。实验结果表明，所提算法能有效分离仿真场景下的双路QPSK、双路16QAM、BPSK+QPSK混合信号以及实测双路BPSK混合信号。在平稳信道的双路QPSK分离中，当误码率为10

-3

时，所提算法相较于深度分离网络（DSNet）、膨胀密集自编码器（DDAEC）和密集连接的多空洞卷积网络（D3Net）分别有约3 dB、2.5 dB和2 dB的性能提升；在非平稳信道的双路QPSK分离中，当误码率为5×10

-3

时，所提算法相较于DSNet、DDAEC和D3Net算法分别有约4 dB、1 dB和0.5 dB的性能提升；在实测双路BPSK的分离中，当误码率为5×10⁻

时，相较于逐留存路径处理算法（PSP）、DSNet、DDAEC和D3Net算法分别有约4 dB、1 dB、0.5 dB和2.5 dB的性能提升。

Abstract

To address the problems of high bit error rates

and high computational complexity in traditional single-channel blind source separation methods

a single-channel blind separation algorithm based on a dual-encoder

-VAE and attention mechanisms

was proposed. A dual-encoder parallel structure was employed to extract signal features

dilated convolutions were introduced to enlarge the receptive field

and a Conformer module was utilized to fuse local and global features. In the decoder

an ECA module was incorporated to improve reconstruction quality

and multi-loss joint optimization was adopted for model training. Experimental results show that the proposed algorithm can effectively separate dual QPSK

dual 16QAM

and BPSK+QPSK mixed signals in simulated scenarios

as well as practical dual BPSK mixed signals. In the separation of dual QPSK signals over stationary channel

the proposed algorithm achieves about 3 dB

2.5 dB and 2 dB performance gains compared with DSNet

DDAEC and D3Net algorithms at a BER of 10⁻³

respectively. While the channel is non-stationary

the proposed algorithm achieves about 4dB

1 dB and 0.5 dB performance gains compared with DSNet

DDAEC and D3Net algorithms at a BER of 5×10⁻

respectively. In the separation of practical dual BPSK signals

the proposed algorithm achieves about 4 dB

1 dB

0.5 dB and 2.5 dB performance gains compared with PSP

DSNet

DDAEC and D3Net algorithms at a BER of 5×10⁻

respectively.

关键词

Keywords

references

邓文 , 黄知涛 , 王翔 . 单通道通信信号盲分离方法的研究进展综述 [J ] . 通信学报 , 2023 , 44 ( 08 ): 179 - 194 .

Deng W , Huang Z T , Wang X . Overview of research progress on blind separation methods forsingle channel communication signal [J ] . Journal on Communications , 2023 , 44 ( 08 ): 179 - 194 .

Dey P , Trivedi N , Satija U , et al . Single channel blind source separation for MISO communication systems [C ] // 2017 IEEE 86th Vehicular Technology Conference (VTC-Fall) . Toronto, Canada , 2017 : 1 - 5 .

栾海妍 , 宋连萍 , 江桦 . 基于能量算子的单通道盲分离中源信号数目估计算法 [J ] . 信息工程大学学报 , 2011 , 12 ( 03 ): 295 - 301 .

Luan H Y , Song L P , Jiang H . Source number estimation algorithm in single channel blind source separation using energy operators [J ] . Journal of Information Engineering University , 2011 , 12 ( 03 ): 295 - 301 .

熊涛 , 崔鹏辉 , 江桦 , 等 . 应用判决反馈的混合信号单通道盲分离算法 [J ] . 信号处理 , 2014 , 30 ( 12 ): 1523 - 1529 .

Xio T , Cui P H , Jiang H , et al . Single channel blind separation of mixed signals applying decision feedback [J ] . Journal of Signal Processing , 2014 , 30 ( 12 ): 1523 - 1529 .

Liu X , Guan Y L . Single-channel blind separation of unsynchronized multiuser PSK signals with non-identical sampling frequency offsets [J ] . IEEE Communications Letters , 2022 , 26 ( 11 ): 2774 - 2778 .

马欢 , 江桦 . 改进的粒子滤波单通道盲分离算法 [J ] . 数据采集与处理 , 2016 , 31 ( 05 ): 1051 - 1058 .

Ma H , Jiang H . Improved single-channel blind separation algorithm based on particle filtering [J ] . Journal of Data Acquisition and Processing , 2016 , 31 ( 05 ): 1051 - 1058 .

Ma H , Zheng X , Yu L , et al . A novel end-to-end deep separation network based on attention mechanism for single channel blind separation in wireless communication [J ] . IET Signal Processing , 2023 , 17 ( 2 ): e12173 .

Pandey A , Wang D . Densely connected neural network with dilated convolutions for real-time speech enhancement in the time domain [C ] // 2020 IEEE International Conference on Acoustics, Speech and Signal Processing(ICASSP) . Barcelona, Spain , 2020 : 6629 - 6633 .

Takahashi N , Mitsufuji Y . D3Net: densely connected multidilated DenseNet for music source separation [J ] . arXiv preprint arXiv: 2010.01733 , 2020 .

Pandey L , Kumar A , Namboodiri V . Monaural audio source separation using variational autoencoders [C ] // Proceedings of the Annual Conference of the International Speech Communication Association (INTERSPEECH) , 2018 : 3489 - 3493 .

Goldberger J , Gordon S , Greenspan H . An efficient image similarity measure based on approximations of KL-divergence between two Gaussian mixtures [C ] // Proceedings of the 9th IEEE International Conference on Computer Vision (ICCV) . Nice, France : IEEE , 2003 : 487 - 493 .

Neri J , Badeau R , Depalle P . Unsupervised blind source separation with variational auto-encoders [C ] // Proceedings of the 29th European Signal Processing Conference (EUSIPCO) . Dublin, Ireland : IEEE , 2021 : 311 - 315 .

Alaghbari K A , Lim H S , Jin B , et al . Source separation in joint communication and radar systems based on unsupervised variational autoencoder [J ] . IEEE Open Journal of Vehicular Technology , 2023 , 5 : 56 - 70 .

Webster M B , Lee J . Blind source separation of single-channel mixtures via multi-encoder autoencoders [J ] . arXiv preprint arXiv: 2309.07138 , 2023 .

Vanessa B , Seljak U . Probabilistic autoencoder [J ] . Transactions on Machine Learning Research , 2022 , 9 : 1 - 25 .

Burgess C P , Higgins I , Pal A , et al . Understanding disentangling in β-VAE [J ] . arXiv preprint arXiv: 1804.03599 , 2018 .

王钧谕 , 高勇 . 基于CNN和Transformer的双路径语音分离 [J ] . 通信技术 , 2023 , 56 ( 05 ): 585 - 589 .

Wang J Y , Gao Y . Dual–path speech separation based on CNN and Transformer . Communications Technology , 2023 , 56 ( 05 ): 585 - 589 .

Wang Q , Wu B , Zhu P , et al . ECA-Net: Efficient channel attention for deep convolutional neural networks [C ] // Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) . Seattle, USA : IEEE , 2020 : 11531 - 11539 .

浏览量

下载量

CSCD

文章被引用时，请邮件提醒。

提交

工具集

关联资源

基于增强关系图卷积网络的数据违规转售检测方法

基于多尺度卷积和通道注意力机制的网络流量异常检测方法

基于CNN-BiLSTM-Attention融合模型的差分隐私轨迹重构攻击

基于深度残差定点网络的太赫兹UM-MIMO系统信道估计算法

MDA-MIM：一种融合多尺度特征与双重注意力机制的雷达回波图预测模型

基于双编码器β-VAE与注意力机制的单通道盲分离算法

Single-channel blind source separation algorithm based on a dual-encoder β-VAE and attention mechanisms

DOI：10.11959/j.issn.1000-436x.TXXB260086

摘要

Abstract

关键词

Keywords

references