人工智能培训

搜索

机器学习论文:基于深度学习的无线电调制分类的数据增强(Data Augmentation for Deep Learning-based Radio Modul

[复制链接]
xujinjin1 发表于 2019-12-9 14:47:52 | 显示全部楼层 |阅读模式
xujinjin1 2019-12-9 14:47:52 219 0 显示全部楼层
机器学习论文:基于深度学习的无线电调制分类的数据增强(Data Augmentation for Deep Learning-based Radio Modulation  Classification)深度学习最近已被用于自动分类接收到的无线电信号的调制类别,而无需人工经验。然而,训练深度学习模型需要大量数据。训练数据不足会导致严重的过拟合问题并降低分类精度。为了应对小型数据集,数据增强已广泛用于图像处理中以扩展数据集并提高深度学习模型的鲁棒性。然而,在无线通信领域,尚未研究不同数据增强方法对无线电调制分类的影响。在本文中,我们通过基于深度学习的最新调制分类器评估不同的数据增强方法。根据调制信号的特性,考虑了三种增强方法,即旋转,翻转和高斯噪声,它们可以应用于深度学习算法的训练阶段和推理阶段。数值结果表明,三种增强方法都可以提高分类的准确性。其中,旋转增强方法的性能优于翻转方法,两者均比高斯噪声方法具有更高的分类精度。在仅训练数据集的12.5%的情况下,联合旋转和翻转增强策略可以获得的分类准确度甚至比没有增强的初始训练数据集的100%更高。此外,通过数据增强,可以使用较短的无线电样本成功地对无线电调制类别进行分类,从而简化了深度学习模型并缩短了分类响应时间。
Deep learning has recently been applied to automatically classify themodulation categories of received radio signals without manual experience.However, training deep learning models requires massive volume of data.Aninsufficient training data will cause serious overfitting problem and degradethe classification accuracy.To cope with small dataset, data augmentation hasbeen widely used in image processing to expand the dataset and improve therobustness of deep learning models.However, in wireless communication areas,the effect of different data augmentation methods on radio modulationclassification has not been studied yet.In this paper, we evaluate differentdata augmentation methods via a state-of-the-art deep learning-based modulationclassifier.Based on the characteristics of modulated signals, threeaugmentation methods are considered, i.e., rotation, flip, and Gaussian noise,which can be applied in both training phase and inference phase of the deeplearning algorithm.Numerical results show that all three augmentation methodscan improve the classification accuracy.Among which, the rotation augmentationmethod outperforms the flip method, both of which achieve higher classificationaccuracy than the Gaussian noise method.Given only 12.5\% of training dataset,a joint rotation and flip augmentation policy can achieve even higherclassification accuracy than the baseline with initial 100\% training datasetwithout augmentation.Furthermore, with data augmentation, radio modulationcategories can be successfully classified using shorter radio samples, leadingto a simplified deep learning model and shorter the classification responsetime.机器学习论文:基于深度学习的无线电调制分类的数据增强(Data Augmentation for Deep Learning-based Radio Modulation  Classification)
URL地址:https://arxiv.org/abs/1912.03026     ----pdf下载地址:https://arxiv.org/pdf/1912.03026    ----机器学习论文:基于深度学习的无线电调制分类的数据增强(Data Augmentation for Deep Learning-based Radio Modulation  Classification)
回复

使用道具 举报

您需要登录后才可以回帖 登录 | 立即注册

本版积分规则 返回列表 发新帖

xujinjin1当前离线
新手上路

查看:219 | 回复:0

快速回复 返回顶部 返回列表