| 摘要 |
To watermark a covertext comprising audio or video signals 10, a two-dimensional noise pattern of predetermined height and width is generated and filtered in one dimension with a cut-off frequency which varies with the position in the pattern. This is used to generate a key 1 which is used by an encoder 2 to watermark the covertext and which can be used subsequently by a decoder 3 to decode the resulting stegotext 15. The key may be used by the encoder to modulate the log power spectrogram 11 obtained by fast Fourier transformation and rectangular polar conversion of the covertext signal. The origin of the covertext can thus be identified in an imperceptible way, and unauthorised reproduction of the covertext may be prevented. |
