Digital Audio Watermarking Utilizing Discrete Wavelet Packet Transform

• Main Authors: Chi-sheng Liu, 劉啟昇
• Other Authors: De-yu Wang
• Format: Others
• Language: zh-TW
• Published: 2004
• Online Access: http://ndltd.ncl.edu.tw/handle/55g4th

碩士 === 朝陽科技大學 === 網路與通訊研究所 === 92 === The past few years have seen an explosion in the use of digital media. Editing the content of digital signals is easy, safe and economical. Thus, the rights of the author are endangered. The digital watermarking technique tries to solve the ownership problems and protect the copyright of the owner. In digital audio watermarking, the goal is to embed some meaningful message information in the original audio signal without perceptional distortion. Up to now, audio watermarking technique in frequency domain mostly takes advantage of Short Time Fourier Transform (STFT) to analyze the auditory masking effect. The STFT estimates spectrum in fixed window length during the whole audio signal. Furthermore, using longer window length obtains more complete frequency resolution certainly than fixed length, but leads to incomplete time resolution. In addition to the masking effect, there are many nonlinear properties in Human Auditory System (HAS). For instance, the resolution of low frequency in ear is more sensitive than high frequency. Therefore, the time-frequency resolutions are constant that can not react to the nonlinear properties. In this paper, we propose Discrete Wavelet Packet Transform (DWPT) to analyze masking effect efficiently. The idea is to employ dynamic and adaptive window length of analysis with the entire audio signal. Thus, the DWPT can not only improve on analysis of the nonstationary audio signal but also fit the frequency resolutions for the nonlinear properties in HAS. After embedding, the watermarked signal is robust enough to common audio processing such as cropping, shifting and MPEG audio compression. In the detection/extraction phase, the characteristic of autocorrelation function like the impulse function is employed to detect the watermark signal is existed or not. If it is, then the meaningful message can be extracted correctly from the watermarked signal without the original signal.