A music watermarking method based on the multi-band power distribution of copyright owner’s speech

Music infringement has emerged as a significant concern and the copyright protection of digital music is becoming more important. Most music copyright protection methods use images as watermark information. Hence, the original watermark needs to be embedded in the music multi-times to achieve copyright protection. It cannot guarantee the accuracy of the embedded watermark. And it may also degrade the quality of the watermarked music. In order to solve that issue, a blind music watermarking is proposed to protect the copyright of music. Comparing with the existing music copyright protection methods, the proposed method generates a watermark from the speech of a copyright owner that can ensure the uniqueness of the watermark. When generating the watermark, the multi-band power distribution of speech recording is calculated first. Subsequently, the discrete cosine transformation is applied to the sub-graphs of the multi-band power distribution, and the watermark is generated using the resulting Discrete Cosine Transformation (DCT) coefficients. After that, the watermark is embedded into the coefficients of Discrete Wavelet Transformation (DWT) using quantization norms. The experiments demonstrate that the Signal to Noise Ratio (SNR) values are higher than 58db between the original music and watermarked music. Meanwhile, the Bit Error Rate (BER) and Normalized Correlation (NC) values are zeros and ones, respectively. In conclusion, the evaluation results show that the proposed method can achieve excellent imperceptibility and robustness. Additionally, it can also effectively resist conventional processing operations such as noise addition, filtering, and so on.