一种基于离散余弦变换的数字语音篡改恢复算法

A Tamper Recovery Scheme for Digital Speech Forensics Based on Discrete Cosine Transform

摘要: 利用离散余弦变换（Discrete Cosine Transform，DCT），提出了一种篡改恢复的语音水印算法。首先，分析了语音信号DCT系数的分布，选取部分DCT系数并编码，生成语音信号压缩数据。然后，将帧号和压缩数据一起作为水印信息，嵌入到语音信号DCT域中，得到含水印的信号。在语音被攻击后，由帧号定位被攻击帧，提取被攻击帧的压缩数据，以此来重构被攻击的内容进行篡改恢复。最后，通过实验证明了本文算法的有效性。

Abstract: A speech watermarking algorithm for tampering recovery based on Discrete Cosine Transform(DCT) is proposed. First, the distribution of DCT coefficients of the speech signal is analyzed, and some DCT coefficients are selected and encoded to generate compressed data of the speech signal. Then, the frame number and compressed data are used as the watermark and embedded in the DCT domain to obtain a watermarked signal. When the speech is attacked, the attacked frame is detected by the frame number. And the compressed data of the attacked frame is extracted to reconstruct the attacked content for tampering recovery. Finally, the experiments are conducted to verify the effectiveness of the algorithm.