基于边际Fisher深度自编码器的电台指纹特征提取

Radio Fingerprint Extraction Based on Marginal Fisher Deep Autoencoder

针对在少量有标签样本条件下传统方法训练不充分而且难以准确提取通信电台指纹特征的问题,文中提出基于边际Fisher深度自编码器的电台指纹特征提取算法.以深度自编码器为基础,训练过程分为无监督预训练、基于边际Fisher映射的有监督训练两部分.首先挖掘海量无标签样本中包含的电台个体类别信息,用于深度自编码器最优参数训练.然后在有标签样本的辅助下对训练参数进行基于边际Fisher映射的有监督精校,提高指纹特征对同类型电台个体的鉴别能力.在多个通信电台数据集上进行的分类识别实验表明,文中算法能在小样本训练条件下有效表达同类型通信电台个体之间的差异.

Aiming at the difficulty in radio fingerprint extraction caused by insufficient traditional training methods with small labeled samples, a deep autoencoder regularized by marginal Fisher analysis algorithm for radio fingerprint extraction is proposed. Based on deep autoencoder, the training procedure is divided into two parts, unsupervised pre-training and supervised finetuning based on marginal Fisher analysis. Firstly, the radio individual class information contained in the large amount of unlabeled samples is extracted. And the information is sent to the deep autoencoder for parameters optimization. Then, the trainable parameters are analyzed on the basis of marginal Fisher method with the assistant of labeled samples to improve the discriminant capability of fingerprint feature between radio individuals of the same model. The classification experiment is conducted on several communication radio signal datasets. The results show that the difference of radio individuals of the same model can be represented effectively bythe proposed algorithm.