Study on Source-filter Model Based Hybrid Speech Watermarking

摘要

As a kind of information carrier, digital speech contains important values. However, illegal use of advanced digital techniques has enabled digital speech to be easily reduplicated and edited, which would cause big problems. Speech watermarking can effectively solve these problems by embedding digital information into the original speech signal. This paper proposes a hybrid watermarking method for speech signal. The hybrid method utilizes the source-filter model of speech production so that the speech signal can be separated into two independent components, i.e., the residue information and the vocal tract information (characterized by formants). The residue and the vocal tract information are separately embedded with watermarks with the quantization index modulation (QIM) based and the formant enhancement (FE) based watermarking. The independence between the residue and vocal tract information ensures that the performance of one watermarking method will not be affected by the other, which makes the hybrid watermarking feasible. Objective evaluations related to inaudibility and robustness were carried out to evaluate the proposed method. The results showed that the proposed method could satisfy inaudibility and the robustness could be increased in comparison with single method by taking advantages of both two methods. These results verified the effectiveness of the proposed method.