关键词: 语音取证, 噪声估计, 篡改检测, 突变点检测

Abstract: Heterologous splicing is a typical tampering behavior for digital voice, which is utilized to change the content and semantic of the target voice. Since the tempered voice is formed by splicing two or more voice segments recorded in different scenes together, it might have different background noises in the spliced parts. This paper proposed a tampering detection algorithm for voice heterologous splicing based on the noise consistency. Firstly, Time-Recursive Averaging (TRA) algorithm is applied to extract the background noise contained in the voice. Then, Change Point Detection (CPD) algorithm is used to detect the abrupt changes of the variance of the extracted noise. Finally, the tampering positions of the testing voice could be located if there exist abrupt changes. Experimental results show that the proposed algorithm can achieve good performance in detecting the heterologous splicing for digital voice.

Key words: speech forensic, noise estimation, tampering detection, change-point detection