Distortion based Watermark Extraction Technique Using 1D CNN
Distortion based Watermark Extraction Technique Using 1D CNN
复制标题
DOI：
10.1109/icaiic51459.2021.9415200
复制
发表时间：
2021-04
期刊：
影响因子：
--
通讯作者：
Yuto Matsunaga;N. Aoki;Y. Dobashi;T. Kojima
中科院分区：
文献类型：
--
作者：
Yuto Matsunaga;N. Aoki;Y. Dobashi;T. Kojima
我们已经提出了一个新的概念，数字水印技术的音乐数据，重点是使用（a）的声音合成和音效技术。先前提出的技术被证实是高通滤波的漏洞。本文描述了传统嵌入技术和采用深度神经网络的改进牵引技术的细节。本文介绍了评估所提出的技术对高通滤波的电阻的实验结果。实验结果表明，本文提出的方法具有较好的抗高通滤波攻击能力（B），而传统方法的抗高通滤波能力较差。
We have proposed a novel concept of a digital watermarking technique for music data that focuses on the use (a) of sound synthesis and sound effect techniques. The previous proposed technique was confirmed a vulnerability to high-pass filtering. This paper describes the details of the conventional embedding technique and the improved traction technique that employs the Deep Neural Networks. This paper describes the experimental results of evaluating the resistance of the proposed technique against high-pass filtering. It is demonstrated that the proposed technique in this paper has appropriate resistance (b) against high-pass filtering attack, which was not good at conventional technique.