| 講演抄録/キーワード |
| 講演名 |
2019-01-11 09:55
Study on Digital Audio Watermarking Method Based on Singular Spectrum Analysis with Automatic Parameter Estimation Using a Convolutional Neural Network ○Kasorn Galajit(JAIST)・Jessada Karnjana(NECTEC)・Pakinee Aimmanee(SIIT)・Masashi Unoki(JAIST) EMM2018-86 |
| 抄録 |
(和) |
(まだ登録されていません) |
| (英) |
An audio watermarking method based on the singular-spectrum analysis (SSA) with a convolutional neural network (CNN) for parameter estimation is studied. A watermark is embedded by modifying a part of a singular spectrum of an original signal. The modification part must be appropriately selected to balance the robustness and sound quality. In our previous work, a differential evolution (DE) algorithm could effectively select the proper part. However, it is time-consuming. In this study, we replaced DE with CNN-based parameter estimation. The experimental results showed that the computational time is significantly reduced by 96,923 times. The sound quality and the robustness were well balanced. Moreover, the scheme was a blind method. |
| キーワード |
(和) |
/ / / / / / / |
| (英) |
automatic parameter estimation / differential evolution / singular spectrum analysis / convolutional neural networks / audio watermarking / / / |
| 文献情報 |
信学技報, vol. 118, no. 382, EMM2018-86, pp. 25-30, 2019年1月. |
| 資料番号 |
EMM2018-86 |
| 発行日 |
2019-01-03 (EMM) |
| ISSN |
Online edition: ISSN 2432-6380 |
著作権に ついて |
技術研究報告に掲載された論文の著作権は電子情報通信学会に帰属します.(許諾番号:10GA0019/12GB0052/13GB0056/17GB0034/18GB0034) |
| PDFダウンロード |
EMM2018-86 |