音訊展頻浮水印檢測與解碼之研究

Abstract

由於網際網路的廣泛應用及影音壓縮技術的躍進，使得盜版傳播的問題日益嚴重，為保障原創者之智慧財產，在多媒體檔案中隱藏認證資訊有其必要性。最佳之浮水印設計必須兼顧隱密性、強軔度及可靠度，隱密性使其不被盜版者察覺，強軔度使之能對抗惡意破壞，而可靠度能減少誤警情況的發生。我們採用離散餘弦轉換配合直序展頻技術來建構嵌入浮水印之函數，主要之關鍵在於設定一離散餘弦係數之廣義高斯機率模型，再根據最大相似度理論，設計最佳的檢測與解碼機制。除此之外，我們也探討非特定機率模型的一般情況，並推導出相關之檢測與解碼演算法。

In this thesis, we address the problem of the performance analysis of audio watermarking systems that use a spread spectrum technique in the discrete cosine transform (DCT) domain. Two tests are involved in the ownership verification process. First, a watermark detector decides whether the audio under test contains a watermark generated with a certain key. If the audio is watermarked , then authorship by the key holder is proved and extraction of hidden message can be performed by a detector. Most current research concentrate on correlation detectors, despite evidence showing that the underlying Gaussian model assumption does not match the intrinsic natures of DCT coefficients. Recognizing this, we first investigate a statistical approach that uses the generalized Gaussian probability to characterize the DCT coefficients and then use it as a basis for the application of statistical decision theory to the design of efficient detector and decoder structures. We also generalize this approach to the possible nonexistence of a statistical description of the original audio.