Audio Digital Watermarking Algorithm Based On Spectrogram

With the rapid development of internet and multimedia technology, the intellectual property rights protection and f the integrity of the content certification become a pressing issue to be settled urgently. Traditional cryptographic techniques can not meet people's needs, the digital watermarking technology arises at the historic moment. Audio digital watermarking technology as a major branch of the digital watermarking, has gradually attracted universal attention. But realization of inaudibility for audios is much more difficult than realizing invisibility for images, the reason for this is that audio signals are represented by much less samples per time interval compared with images and video. Furthermore, the human auditory system (HAS) is much more sensitive than the human visual system (HVS).All along, the image digital watermarking and audio digital watermarking research is independent of each other, and the research capacity has become the bottleneck for audio digital watermarking. In this paper, it is the first step to research the relations between the audio signals and spectrogram. This will be a kind of new method in audio digital watermarking. Major original works in the research are listed in details as follows:1. The relationship between the voice signals and the spectrogram is studied. Through theoretical analysis, voice signals in time-domain and the spectrogram can be transformed each other. Through four different experiments, results showed that the similarity between the original and the signal converted from the spectrogram are 1.0000.2. The theory of using the spectrogram into the audio digital watermarking is proposed. First of all, the relationship among the amplitude spectrum,the energy spectrum and the spectrum is researched,and a mathematical model to prove the theory is build. It is proved theoretically that the spectrogram in the band for watermark embedding arising from the magnitude of the signal distortion, is less than directly in the amplitude spectrum and the energy spectrum for watermark embedding directly caused by the magnitude of the signal distortion.3. Algorithm of the audio digital watermarking based on the space-domain of spectrogram is proposed. In the algorithm, the watermark signals are directly embedded in the Middle Frequency Band of the spectrogram with the image watermark embedding algorithm, secondly, the spectrogram embedded watermark signals are returned to the signal in time-domain. Experimental results show that the watermark embedding algorithm can effectively resist shear attacks.4. Algorithm of the audio digital watermarking based on the frequency-domain of spectrum is proposed. It is to select the appropriate matrix in the spectrum, the digital watermark algorithm in FFT domain and digital watermark in DCT domain based on the matrix .To sum up the above arguments, this paper uses the spectrum in audio digital watermark by researching the relationship between the audio signal and spectrogram , through the spectrogram, a bridge between the audio signal and the image digital watermarking has been built. It is a successful application of side information and confusion network in audio digital watermarking .