基于时序卷积网络的信道编码闭集识别

Channel coding closed set recognition based on temporal convolutional network

针对传统识别算法对信号的先验知识要求较高、人工特征提取复杂、低信噪比环境下识别率较低等问题,提出了一种基于时序卷积网络(TCN)的卷积码参数识别方法.引入了深度学习算法处理盲识别问题,依据卷积码的马尔可夫性,将码字作为时间序列处理,把已知类型的编码序列作为时序卷积网络模型的输入进行监督学习,根据训练好的模型对接收端接收到的未知编码信号进行闭集识别分类.实验结果表明:当信噪比大于5 dB时,单一参数类型与混合参数类型平均识别准确率分别大于99.60%和99.50%,且在相关算法对比中有较好的识别表现.

In order to solve the problems that traditional recognition algorithms require high prior knowledge of signals,complex artificial feature extraction,and low recognition rate in low SNR environment,a parameter recognition method of convolutional codes based on temporal convolutional network(TCN) was proposed.The method was introduced into deep learning algorithm to tackle the problem of blind identification based on Markov of convolution code. The code words were processed as time series.The known types of coding sequence was used as the input sequence convolution network model for supervised learning,and the unknown signals received by the receiver were classified according to the trained model.The experimental results show that when the SNR is more than 5 dB,the average recognition accuracy of single parameter type and mixed parameter type is more than 99.60%and 99.50% respectively,and the recognition performance is better in the correlation algorithm comparison.