| Summary: | 碩士 === 元智大學 === 通訊工程學系 === 98 === Digital fingerprinting is a technique to protect intellectual properties. Illegal distribution of the digital media can be prevented. If digital fingerprints embedded in the media file can be extracted,and the illegal distributors can be found. Ideally,digital fingerprints should be imperceptible and can resist collusion attacks. The colluder may combine several copies of media files,with the intention of scrambling the digital fingerprints in the files to prevent traiter tracing.In this research, we explore methods to improve the imperceptibility of digital fingerprints that can resist collusion attacks.
In Cha and Kuo’s 2007 research, a technique similar to spread spectrum is proposed,where the Hadamard sequences were used as codewords for digital fingerprints in digital music. We believe that the imperceptibility of digital fingerprints is related to the peak-to-average power ratio (PAPR) of the codewords in time and frequency domains.Codewords with samller PAPR are less perceptible. We compare the PAPR of several well known orthogonal sequence sets,e.q. Hadamard sequences, Golay Hadamard sequences,inverse discrete Fourier transform sequences,and cyclicly-shifted pseudo-noise sequences. The PAPR of Golay Hadamard sequences are the lowest.
Since the Hadamard sequences used in Cha and Kuo’s 2007 research have high PAPR and not ideal imperceptibility, we propose to use orthogonal Golay Hadamard sequences as codewords. It’s low PAPR in time and frequency domains allow us to create digital fingerprints whose strengths can be adjusted to follow the strengths of the music signal, so that the imperceptibility can be improved.The cyclicly-shifted pseudo-noise sequences often used in communication systems also have low PAPR in both time and frequency domain. However, since these sequences can have high cross crrelations after being shifted,the resistance towards collusion attacks would be affected.
|
|---|
