<title>Robust audio watermarking for copyright protection</title>

Improved watermark detection for spread-spectrum based watermarking using independent component analysis

“…The proposed ICAWD can be used to detect watermark for most of the existing SS-based watermark embedding schemes [1][2][3]. The simulation results presented here are obtained using our frequency selective spread spectrum (FSSS) audio watermarking scheme [3].…”

“…In general, the existing watermark detectors can be classified into two categories: (a) informed detectors that assume that the host signal is available at the detector during watermark detection process, and (b) blind detectors that assume that the host signal is not available at the detector for watermark detection. Similarly, watermark embedding schemes can be classified into two major groups: (a) blind embedding implies that the watermark embedder does not use the host signal information during the watermark embedding process (spread spectrum based watermarking schemes [1,2,3,6] fall in this category), and (b) informed embedding implies that the watermark embedder exploits the properties of the host media during the embedding process (quantization index modulation based watermarking schemes [1,4,6] belong to this category).…”

A new adaptive digital audio watermarking based on support vector machine

“…Seok et al (2002) gave an audio watermarking scheme by exploiting the human perceptual characteristics of the audio signal to regulate the embedding strength, but it is not very robust to some audio signal processing such as re-sampling, requantization and compression. The current self-synchronization algorithm cannot extract feature points steadily, besides, it usually need large number of threshold values which make it more difficult to be applied (Wu et al, 1999;Li and Xue, 2003a). Girin and Marchand (2004) proposed a speech signal watermarking using the sinusoidal model and amplitudes, phases and digital frequencies modulation of the partials.…”

A robust digital audio watermarking based on statistics characteristics

“…Among them, all-list-search strategy need great calculating amount and has high false positive rate [6]; most second strategy cannot achieve blind detection [5][6][7], Seok et al [8] gave an audio watermarking scheme by exploiting the human perceptual characteristics of the audio signal to regulate the embedding strength, but it is not very robust to some audio signal processing such as re-sampling, re-quantization and compression; the current selfsynchronization algorithm cannot extract feature points steadily, besides, it usually need large number of threshold values which make it more difficult to be applied [9,10], Girin et al [11] proposed a speech signal watermarking using the sinusoidal model and amplitudes, phases and digital frequencies modulation of the partials, Hee et al [12] proposed an audio watermarking algorithm through modification of tonal maskers, but these methods suffer from poor robustness against time-scale modification and pitch shifting. By contrast, synchronization code strategy has more obvious technological advantages.…”