LSTM神经网络下多噪声环境通信弱信号识别仿真  

Simulation of Weak Signal Recognition for Multi-Noise Environmental Communication under LSTM Neural Network

作　　者：石娟娟[1] 黄越嘉[1] SHI Juan-juan;HUANG Yue-jia(South China Agricultural University,Guangzhou Guangdong 510000,China)

机构地区：[1]华南农业大学,广东广州510000

出　　处：《计算机仿真》2024年第10期159-163,共5页Computer Simulation

基　　金：2024年度江门市科学规划课题:原创音乐作品赋能乡村文化振兴的实践路径研究——以“绿美乡村”梯田音乐会为例的研究成果2024JMZG128;2024年度江门市科学规划课题:基于“元宇宙”概念下的广东省数字乡村文化振兴研究与实践:以江门百合镇为例的研究成果2024JMZG030;2023年国家级大学生创新创业训练计划项目《基于多维视角的客家非遗文化与动画交互融合研究——以元善镇香火龙为例》(课题编号:202310564072)。

摘　　要：由于在多噪声环境下,微弱信号容易被噪声掩盖,导致通信弱信号识别难度较大且极耗费时间。为此,提出一种基于LSTM神经网络的多噪声环境通信弱信号识别方法。通过贪婪权值算法(Takenaka-Malmquist Greedy Algorithm,TMGW)压缩通信弱信号,获取通信弱信号和噪声在时频面上的能量分布特点,对其降噪处理。建立长短记忆(Long Short Term Memory,LSTM)神经网络预测模型结构图,通过损失函数优化训练的权重参数,获取网络训练结果,计算实际值与预测值之差,作为高斯分布检测模型的特征值,实现复杂噪声下通信弱信号识别。仿真结果表明,采用所提方法可以有效地识别通信弱信号,全面提升多噪声环境通信弱信号识别效率。Generally,weak signals are easily masked by noise,so it is difficult and time-consuming to identify weak signals in communication.Therefore,a method of identifying weak signals in communication in a multi-noise environment based on LSTM neural network was proposed.The Takenaka-Malmquist Greedy Algorithm(TMGW)was adopted to compress weak communication signals,thus obtaining the energy distribution characteristics of weak signals and noise on the time-frequency plane and denoising the signals.Then,a structure diagram of the Long Short-Term Memory(LSTM)neural network prediction model was constructed.Meanwhile,the trained weight parameters were optimized by the loss function,and then network training results were obtained.Furthermore,the difference between the actual value and the predicted value was calculated as the characteristic value of the Gaussian distribution detection model.Thus,the recognition of weak communication signals in complex noise is realized.Simulation results prove that the proposed method can effectively identify weak signals and comprehensively improve the recognition efficiency of weak signals in a multi-noise environment.

关 键 词：神经网络 多噪声环境 通信弱信号 识别 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]