Speech watermarking based tamper detection and recovery scheme with high tolerable tamper rate

Speech watermarking has been widely used for tamper detection and recovery. In this paper, we propose a speech watermarking based tamper detection and recovery method after analyzing the characteristics of continuous and discontinuous tamper. The authentication watermarks and recovery watermarks are embedded into the original speech using align embedding and misalign embedding strategies, respectively. In particular, the misalign embedding strategy which distributes the recovery watermarks repeatedly and widely can effectively prevent the speech segment and its recovery watermarks from being tampered simultaneously, which significantly increases the tolerable tamper rate (TTR) of the proposed method. Several experiments concerning inaudibility, recovery rate, sound quality of recovered speech, and recovery percentage were carried out to evaluate the proposed method. The obtained results suggested that the proposed method had good inaudibility. Moreover, it could tolerate high tamper rate (around 50%) and provide satisfactory recovery rate (100%) and speech quality (PESQ≥ 3.0 ODG and LSD ≤ 1.0 dB) under continuous tamper (for N ≥ 6). Similarly, it could recovery most of the speech after discontinuous tamper even under high tamper rate. These results verified the effectiveness of the proposed method.