Motivation:
Speech signal is an important information carrier in many social applications such as WeChat and GoogleTalk；
Modern digital technologies have put the security of speech at risk.
Solution: Watermarking is a promising solution to protect the speech signals by embedding digital data into them [1, 2].
Problem:
Many existing methods cannot satisfy the requirements of watermarking, e.g., inaudibility and robustness, simultaneously;
It is easy to extract the watermarks in ideal situations (without noise/ interferences), but when there are noise/interferences, the embedded watermarks will fail to extract.
Aim: Realize a watermarking method based on RPCA and formantmanipulations that can satisfy both inaudibility and robustness.