A novel spread spectrum algorithm for audio watermarking based on wavelet transform

الملخص EN

This paper presents a new wavelet-based spread spectrum audio watermarking algorithm.

In this algorithm, new techniques are proposed.

The first technique is used to specify periods of audio signal to insert a watermark into these periods.

Selected periods must satisfy conditions designed to enable robust and inaudible embedding.

The technique used for that purpose is called Block Selection Process (BSP).

Another technique is proposed for watermark embedding.

In this technique, embedding operation is achieved by generating signal-dependent watermark patterns and adding them to the signal in wavelet domain.

For each watermark bit two patterns are generated; one to be embedded in approximation part and the other to be embedded in details part.

Before adding the patterns into audio signal, patterns are modulated and filtered to reduce a possible distortion might be resulted from these patterns.

In the detection stage, we must locate where the watermarks were embedded before start extracting them.

BSP is used again to accomplish this task.

Then, the watermark patterns are regenerated and correlated with the audio signal to decide whether the watermark exists or not.

In the proposed algorithm, the watermark patterns are regenerated from watermarked audio signal and reshaped before correlation.

A sort of whitening operation is used to enhance watermark detection.

The proposed algorithm has been subjected to and survived robustness -tests and the results are reported.