A SILK Adaptive Steganographic Scheme Based on Minimizing Distortion in Pitch Domain

摘要

SILK, as an efficient speech codec, has been widely used in real-time interactive communication and can be a good carrier for hiding information. This paper proposes a secure steganographic scheme in SILK pitch domain based on speech signal correlation and statistical distribution of pitch lags. Considering that even minor modification of pitch lags will reduce time correlation of speech signal and the statistical characteristics will be disturbed, this paper elaborately designs a distortion function to measure the loss of signal correlation and statistical features introduced by embedding operation. Using syndrome-trellis codes, this paper realizes a SILK adaptive steganography algorithm based on a minimum distortion framework in pitch domain. Experiments show that the proposed scheme can maintain good voice quality and resist the detection of existing steganalysis schemes.