本論文主要針對應用於音訊上的浮水印，提出一套不僅具有良好音訊品質且經音訊攻擊後仍具有高強韌性，且在萃取端並不需要原始音訊的數位浮水印技術。 本文提出的方法，以小波封包 ( Wavelet Packet ) 分頻方式，將樂音訊號經由濾波器群組分成29個次頻帶，其頻寬分布與人類聽覺的26個關鍵頻帶 ( Critical Band ) 相近，藉以找出人耳聲學模型 ( Psychoacoustic Model ) 中的最小遮蔽臨界值 ( Minimum Masking Threshold )，此值將作為嵌入浮水印時更動小波係數的依據。浮水印嵌入時選擇人耳較不敏感的偏低頻係數作為嵌入之用，並且使用區塊分類來過濾不適合嵌入浮水印的係數區塊，將數位浮水印嵌入較不易改變極性之區塊以確保浮水印的強韌性。另一個特點為在浮水印嵌入過程中會產生一組與原始音訊相關的安全鑰匙 ( Secure Key )，萃取端藉由此安全鑰匙即可完整取出嵌入的浮水印資訊，所以並無傳統做法中對原始音訊的需求，可以省去原始音訊的儲存空間。此外為了克服音訊同步的問題，也加入了同步碼搜尋的機制。實驗結果顯示對於MP3 48 Kbps以上之壓縮，以及StirMark各式的攻擊均有足夠的強健性，對於音訊裁切攻擊也能正確的找回同步點位置。 In this thesis, we propose an audio watermarking system based on wavelet packet decomposition and psycho-acoustic modeling. The audio watermarking system can deliver perceptual transparent audio quality, and it is robust against various signal processing or malicious attacks. The original audio signal is first segmented and divided into 29 subbands via wavelet packet decomposition. The bandwidth allocation of the subband decomposition structure is close to the critical band structure of human auditory system. Middle and middle-low subbands are chosen for watermark embedding. A selective embedding method is used to embed watermark into those coefficient blocks with clear block polarities. Modification of selected coefficients is based on the minimum masking threshold of psycho-acoustic model. Instead of the original Audio signal, a secure key is used in the watermark extraction to indicate the locations where watermark bits are embedded. In addition, a synchronization mechanism is imposed to search for the frame boundary of attacked audio signal. Simulation results show that the watermarking system is robust against MP3 compression at 48 Kbps and above. It also survives StirMark attacks for audio and cropping attacks.
