基于CNN的非线性传递函数建模与波形重建  

Nonlinear Transfer Function Modeling And Waveform Reconstruction Based on CNN

作　　者：向前 

机构地区：[1]武昌工学院信息工程学院,湖北武汉430065

出　　处：《数字技术与应用》2017年第9期57-59,61,共4页Digital Technology & Application

基　　金：武昌工学院博士启动基金项目(No.2017BSJ02)

摘　　要：在噪声有源控制、电磁屏蔽效能评估等问题中,对信号传递函数的准确建模是实现控制性能的关键因素。卷积神经网络(CNN)可以多层次提取和组合复杂对象的特征,并在图像和语音等领域取得了广泛应用。本文提出了一种基于CNN模型的非线性通道传递函数建模及输出波形重建方法,并与自适应辨识算法(Fx LMS)进行了比较。仿真试验表明,对于典型的非线性传输通道,基于CNN的传递函数建模和波形重建方法具有更高的精度和可接受的计算时间。In some fields such as active noise control （ANC） system, evaluation of EMP shielding effection and so on, it is of great importance to model signal transfer path function accurately. Convolution neural network（CNN） can extract and combine multilevel features of a complex object, and has been successfully applied to many fields such as image and speech recognition. This paper proposed an algorithm based on CNN to carry out nonlinear transfer function modeling and waveform reconstruction, and compared with the adaptive active noise control algorithm（FxLMS）. The resul of simulation experiment shows, for a typical nonlinear transmission channel, the transfer function modeling and waveform reconstruction based on CNN has a higher accuracy and acceptable computation time.

关 键 词：传递函数建模 CNN 非线性 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]