基于線性預(yù)測倒譜濾波的音頻水印檢測技術(shù)研究
發(fā)布時間：2018-07-17 01:48
【摘要】：隨著網(wǎng)絡(luò)與多媒體技術(shù)的迅速發(fā)展,多媒體產(chǎn)品面臨被輕易地復(fù)制、傳輸與修改的難題。加強媒體信息安全與保護媒體知識產(chǎn)權(quán)的需求日益迫切。數(shù)字水印是維護多媒體產(chǎn)品安全的一種有效解決方案,通過在媒體數(shù)據(jù)中添加代表版權(quán)所有權(quán)的隱秘信息,可以達到保護版權(quán)、鑒別數(shù)據(jù)真?zhèn)渭皹俗R產(chǎn)品等目的。在感知靈敏區(qū)域內(nèi),相比于視覺系統(tǒng),人的聽覺系統(tǒng)具有更高的靈敏度,因此,在數(shù)字音頻信號中添加水印的技術(shù)比靜態(tài)圖像水印技術(shù)的難度更大。傳統(tǒng)的擴頻音頻水印技術(shù)中,宿主信號在水印檢測過程中充當著噪聲源的角色,使得水印檢測的誤碼率比較高。本文的研究內(nèi)容就是為了降低水印檢測誤碼率,同時保證音頻的感知透明性,降低宿主信號對水印的影響。在音頻水印技術(shù)領(lǐng)域,普遍采用減少信號方差來提高水印檢測率。傳統(tǒng)方法分別是Savitzky-Golay濾波、譜包絡(luò)濾波、倒譜濾波和線性預(yù)測。盡管這四種方法都可以減少信號方差,但目前尚缺乏公開的文獻資料分析這四種技術(shù),缺乏實驗代碼實現(xiàn)和實驗分析,因此對各項參數(shù)對實驗性能的影響缺乏定量的數(shù)據(jù)分析和指導(dǎo)。本文通過Matlab實現(xiàn)了這四種降低信號方差方法的實驗研究以及性能分析,并在此基礎(chǔ)上提出了線性預(yù)測倒譜濾波方法,能夠顯著降低樣本點幅度,進而降低信號方差。然后,本文實現(xiàn)了基于擴頻技術(shù)的音頻水印算法的優(yōu)化,在擴頻水印嵌入模塊增加了音頻信號的頻域特性分析,用來平衡水印的嵌入強度,以此保證嵌入水印后的音頻信號具有良好的感知透明性,并將降低信號方差的方法應(yīng)用到音頻水印檢測系統(tǒng)中,降低宿主信號對水印檢測的影響。最后,將提出的擴頻水印優(yōu)化算法(OSS)與Malvar和Florencio提出的改進擴頻水印方法(ISS)進行了感知透明性與誤碼率的比較,并利用Matlab實驗,驗證了優(yōu)化算法可以有效地降低信號對嵌入水印的影響,提高水印檢測性能,相對于ISS,優(yōu)化算法可以降低2%的誤碼率。
[Abstract]:With the rapid development of network and multimedia technology, multimedia products are faced with the problem of easy replication, transmission and modification. The need to strengthen media information security and protect media intellectual property is increasingly urgent. Digital watermarking is an effective solution to maintain the security of multimedia products. By adding secret information representing copyright ownership in media data, we can protect copyright, identify the authenticity of data and identify products. In perceptual sensitive region, human auditory system has higher sensitivity than visual system, so it is more difficult to add watermark to digital audio signal than to static image watermarking. In the traditional spread spectrum audio watermarking technology, the host signal acts as the noise source in the watermark detection process, which makes the bit error rate of watermark detection higher. The purpose of this paper is to reduce the bit error rate of watermark detection, at the same time to ensure the perceptual transparency of audio and reduce the influence of host signal on watermark. In the field of audio watermarking, reducing signal variance is widely used to improve watermark detection rate. The traditional methods are Savitzky-Golay filtering, spectral envelope filtering, cepstrum filtering and linear prediction. Although all four methods can reduce the variance of signals, there is still a lack of open literature and analysis of these four techniques, lack of experimental code implementation and experimental analysis. Therefore, there is a lack of quantitative data analysis and guidance on the influence of various parameters on the experimental performance. In this paper, the experimental research and performance analysis of these four methods for reducing signal variance are realized by Matlab. Based on this, a linear predictive cepstrum filtering method is proposed, which can significantly reduce the sample amplitude and then reduce the signal variance. Then, the audio watermarking algorithm based on spread spectrum technology is optimized, and the frequency domain characteristic analysis of audio signal is added to the spread spectrum watermark embedding module to balance the embedding intensity of the watermark. The method of reducing signal variance is applied to audio watermark detection system to reduce the influence of host signal on watermark detection. Finally, the proposed spread spectrum watermarking optimization algorithm (OSS) is compared with the improved spread spectrum watermarking algorithm (ISS) proposed by Malvar and Florencio. It is verified that the optimized algorithm can effectively reduce the influence of the signal on the embedded watermark and improve the watermark detection performance. Compared with ISS, the optimized algorithm can reduce the bit error rate by 2%.
【學(xué)位授予單位】：華中師范大學(xué)
【學(xué)位級別】：碩士
【學(xué)位授予年份】：2017
【分類號】：TP309.7
本文編號：2128491
[Abstract]:With the rapid development of network and multimedia technology, multimedia products are faced with the problem of easy replication, transmission and modification. The need to strengthen media information security and protect media intellectual property is increasingly urgent. Digital watermarking is an effective solution to maintain the security of multimedia products. By adding secret information representing copyright ownership in media data, we can protect copyright, identify the authenticity of data and identify products. In perceptual sensitive region, human auditory system has higher sensitivity than visual system, so it is more difficult to add watermark to digital audio signal than to static image watermarking. In the traditional spread spectrum audio watermarking technology, the host signal acts as the noise source in the watermark detection process, which makes the bit error rate of watermark detection higher. The purpose of this paper is to reduce the bit error rate of watermark detection, at the same time to ensure the perceptual transparency of audio and reduce the influence of host signal on watermark. In the field of audio watermarking, reducing signal variance is widely used to improve watermark detection rate. The traditional methods are Savitzky-Golay filtering, spectral envelope filtering, cepstrum filtering and linear prediction. Although all four methods can reduce the variance of signals, there is still a lack of open literature and analysis of these four techniques, lack of experimental code implementation and experimental analysis. Therefore, there is a lack of quantitative data analysis and guidance on the influence of various parameters on the experimental performance. In this paper, the experimental research and performance analysis of these four methods for reducing signal variance are realized by Matlab. Based on this, a linear predictive cepstrum filtering method is proposed, which can significantly reduce the sample amplitude and then reduce the signal variance. Then, the audio watermarking algorithm based on spread spectrum technology is optimized, and the frequency domain characteristic analysis of audio signal is added to the spread spectrum watermark embedding module to balance the embedding intensity of the watermark. The method of reducing signal variance is applied to audio watermark detection system to reduce the influence of host signal on watermark detection. Finally, the proposed spread spectrum watermarking optimization algorithm (OSS) is compared with the improved spread spectrum watermarking algorithm (ISS) proposed by Malvar and Florencio. It is verified that the optimized algorithm can effectively reduce the influence of the signal on the embedded watermark and improve the watermark detection performance. Compared with ISS, the optimized algorithm can reduce the bit error rate by 2%.
【學(xué)位授予單位】：華中師范大學(xué)
【學(xué)位級別】：碩士
【學(xué)位授予年份】：2017
【分類號】：TP309.7
【參考文獻】
相關(guān)期刊論文 前10條
1 陳雪松;李昊天;賈瑞成;孫立娜;卜廣龍;;一種基于混沌加密的DWT數(shù)字音頻水印算法[J];計算機與現(xiàn)代化;2014年08期
2 馬艷;李學(xué)偉;王曉艷;;基于ACR的多屏同步互動應(yīng)用技術(shù)淺析[J];電視技術(shù);2013年21期
3 趙學(xué)敏;郭宇弘;鄒學(xué)強;劉建;顏永紅;;用于版權(quán)管理的數(shù)字音頻水印算法[J];電子與信息學(xué)報;2011年10期
4 陳雪松;申美玲;;基于Lorenz混沌加密的DWT音頻水印算法[J];科學(xué)技術(shù)與工程;2011年07期
5 蔡天凈;唐瀚;;Savitzky-Golay平滑濾波器的最小二乘擬合原理綜述[J];數(shù)字通信;2011年01期
6 龔翱;李曾妍;陳剛;;數(shù)字水印技術(shù)及應(yīng)用[J];計算機與信息技術(shù);2010年10期
7 陳衛(wèi)東;王曉亞;解靜;;基于LPCC的多語種識別算法[J];無線電工程;2009年09期
8 劉穎;張萍;丁慶生;;音頻編解碼算法的客觀測試[J];現(xiàn)代電子技術(shù);2006年10期
9 尹浩,林闖,邱鋒,丁嶸;數(shù)字水印技術(shù)綜述[J];計算機研究與發(fā)展;2005年07期
10 梁娟,朱忠軍;基于心理聲學(xué)模型的自適應(yīng)音頻數(shù)字水印算法[J];現(xiàn)代電子技術(shù);2004年20期
相關(guān)博士學(xué)位論文 前1條
1 李力利;基于擴頻技術(shù)的時域音頻水印的研究與改進[D];上海交通大學(xué);2009年
相關(guān)碩士學(xué)位論文 前2條
1 王曉盼;基于擴頻的數(shù)字音頻水印技術(shù)的研究[D];河北科技大學(xué);2015年
2 孫瀅;數(shù)字音頻水印技術(shù)的研究[D];哈爾濱工程大學(xué);2007年
，本文編號：2128491
本文鏈接：http://sikaile.net/falvlunwen/zhishichanquanfa/2128491.html
最近更新
教材專著
