《電子技術(shù)應(yīng)用》
您所在的位置:首頁(yè) > 通信與網(wǎng)絡(luò) > 設(shè)計(jì)應(yīng)用 > 一種結(jié)合雙閾值與WSN的OFDM輻射源個(gè)體識(shí)別
一種結(jié)合雙閾值與WSN的OFDM輻射源個(gè)體識(shí)別
網(wǎng)絡(luò)安全與數(shù)據(jù)治理
劉高輝, 鄭文文
西安理工大學(xué)自動(dòng)化與信息工程學(xué)院
摘要: 針對(duì)在低信噪比情況下OFDM輻射源識(shí)別率低的問題,提出一種雙閾值與小波散射網(wǎng)絡(luò)(Wavelet Scattering Network,WSN)結(jié)合的OFDM信號(hào)時(shí)頻圖輻射源個(gè)體識(shí)別方法。首先,建立包含指紋特征的OFDM信號(hào)指紋模型;其次,利用小波變換對(duì)一個(gè)符號(hào)周期內(nèi)的OFDM信號(hào)進(jìn)行時(shí)頻分析得到時(shí)頻圖;再次,設(shè)計(jì)一種雙參數(shù)閾值函數(shù)模型實(shí)現(xiàn)自適應(yīng)抑制噪聲干擾,提高時(shí)頻圖的圖像質(zhì)量。經(jīng)WSN處理后,從優(yōu)化后的時(shí)頻圖中提取小波散射系數(shù)作為特征集;最后,采用ResNet18進(jìn)行分類識(shí)別。仿真實(shí)驗(yàn)結(jié)果表明,該方法能夠顯著抑制噪聲干擾,在信噪比為-4 dB的條件下,識(shí)別精度達(dá)到87.5%,相較于其他方法表現(xiàn)出更高的識(shí)別精度和抗噪性能。
中圖分類號(hào):TN911文獻(xiàn)標(biāo)識(shí)碼:ADOI:10.19358/j.issn.2097-1788.2025.03.007
引用格式:劉高輝, 鄭文文. 一種結(jié)合雙閾值與WSN的OFDM輻射源個(gè)體識(shí)別[J].網(wǎng)絡(luò)安全與數(shù)據(jù)治理,2025,44(3):39-46.
An OFDM radiation source individual identification combining dual thresholding and WSN
Liu Gaohui, Zheng Wenwen
School of Automation and Information Engineering,Xi′an University of Technology
Abstract: Aiming at the problem of low recognition rate of OFDM radiation source under low signal-to-noise ratio, a dual-threshold and wavelet scattering network (WSN)-based method is proposed for individual recognition of the radiation source in the time-frequency diagram of OFDM signal. Firstly, according to the working principle of OFDM transmitter, the fingerprint model of OFDM signal containing fingerprint features is established; secondly, the time-frequency diagram is obtained by time-frequency analysis of OFDM signal in one symbol period using wavelet transform. Again, a two-parameter threshold function model is designed to realize adaptive suppression of noise interference and improve the image quality of the time-frequency diagram. After WSN processing, wavelet scattering coefficients are extracted from the optimized time-frequency diagram as the feature set; finally, ResNet18 is used for classification and identification. The simulation results show that the method can significantly suppress noise interference, with recognition accuracy reaching 87.5% at a signal-to-noise ratio of -4 dB, outperforming other methods in terms of recognition accuracy and noise-resistant performance.
Key words : time-frequency diagram; dual threshold function; wavelet scattering network; individual identification of OFDM radiation sources

引言

通信輻射源個(gè)體識(shí)別技術(shù)專注于對(duì)通信發(fā)射機(jī)信號(hào)進(jìn)行處理,實(shí)現(xiàn)對(duì)不同通信輻射源個(gè)體的有效區(qū)分[1]。近年來(lái),時(shí)頻分析方法在信號(hào)特征提取與重構(gòu)中展現(xiàn)出顯著優(yōu)勢(shì),已成為現(xiàn)代信息處理的主流方法。

常用的時(shí)頻分析方法涵蓋了短時(shí)傅里葉變換(Short-Time Fourier Transform, STFT)[2-3]、小波變換(Wavelet Transform, WT)[4]和Wigner-Ville分布(Wigner-Ville Distribution, WVD)[5-6]等線性及二次型分布方法。其中,STFT通過固定窗口對(duì)信號(hào)進(jìn)行分段分析,但由于Heisenberg測(cè)不準(zhǔn)原理[7-8],其在分辨率上存在局限性。WVD是一種二次型分布,在處理多分量信號(hào)時(shí),WVD可能會(huì)產(chǎn)生交叉干擾,進(jìn)而對(duì)分析結(jié)果造成干擾。相比之下,WT憑借其時(shí)頻局部性分析的能力,能夠根據(jù)信號(hào)頻率動(dòng)態(tài)調(diào)整時(shí)頻窗的大小,通過平移和伸縮操作對(duì)信號(hào)進(jìn)行多尺度解析?;赪T的優(yōu)越性能,Mallat提出了小波散射網(wǎng)絡(luò)(Wavelet Scattering Network,WSN) [9],該網(wǎng)絡(luò)通過多級(jí)小波分解實(shí)現(xiàn)信號(hào)的多尺度分析,有效減少對(duì)特定小波基的依賴,從而在提取信號(hào)的關(guān)鍵特征和細(xì)節(jié)方面表現(xiàn)出色。WSN不僅具有高度的魯棒性,還能有效保留輸入信號(hào)的穩(wěn)定特征。在結(jié)構(gòu)上,WSN與卷積神經(jīng)網(wǎng)絡(luò)(Convolutional Neural Network, CNN)相似,同樣具備平移不變性和變形穩(wěn)定性,并能夠保留高頻信息進(jìn)行分類[10]。文獻(xiàn)[11]利用具有尺度不變性和穩(wěn)定性的小波散射變換進(jìn)行多普勒信號(hào)分類,取得了顯著成效。文獻(xiàn)[12]則提出了一種基于小波散射網(wǎng)絡(luò)的新型藍(lán)牙信號(hào)射頻指紋識(shí)別方法。文獻(xiàn)[13]提出了一種利用WSN提取射頻指紋特征的方法,通過對(duì)信號(hào)進(jìn)行多尺度分析,提高了特征表示能力,但在低信噪比條件下,其性能仍可能受到噪聲的干擾。針對(duì)這一問題,文獻(xiàn)[14]采用雙參數(shù)閾值函數(shù)模型對(duì)高頻子帶實(shí)現(xiàn)噪聲抑制,從而得到理想的小波降噪效果,不僅有效提升了小波散射網(wǎng)絡(luò)的抗噪性能,還進(jìn)一步增強(qiáng)了其在低信噪比環(huán)境下的特征提取能力。

為解決低信噪比(Signal-to-Noise Ratio,SNR)下輻射源識(shí)別率低的問題,本文提出雙閾值與WSN結(jié)合的OFDM信號(hào)時(shí)頻圖識(shí)別方法。首先,WT轉(zhuǎn)換OFDM信號(hào)為二維時(shí)頻圖,直觀表征時(shí)頻特征;然后,設(shè)計(jì)雙參數(shù)閾值函數(shù)模型自適應(yīng)抑制噪聲,確保特征保留;接著,選擇小波函數(shù)與窗函數(shù)優(yōu)化WSN性能,從優(yōu)化后的時(shí)頻圖深度提取小波散射系數(shù)特征集,高魯棒性表征時(shí)頻細(xì)節(jié);最后,特征輸入ResNet18分類模型,實(shí)現(xiàn)輻射源個(gè)體識(shí)別,改善低SNR下識(shí)別效果不佳問題。


本文詳細(xì)內(nèi)容請(qǐng)下載:

http://ihrv.cn/resource/share/2000006376


作者信息:

劉高輝, 鄭文文

(西安理工大學(xué)自動(dòng)化與信息工程學(xué)院,陜西西安710048)


Magazine.Subscription.jpg

此內(nèi)容為AET網(wǎng)站原創(chuàng),未經(jīng)授權(quán)禁止轉(zhuǎn)載。