结合密钥和随机标准正交基的音频伪装方案

Audio Camouflage Scheme Combining Key and Random Orthonormal Basis

传统基于Tangram的音频伪装方法所采用的变换模型为仿射变换模型,变换精度低且不满足基本的正交关系,从而无法保证秘密音频与公开音频之间的拟合精度,同时当分段变换音频为恒值序列时,需添加随机扰动以保证变换后音频的恢复质量,由此会降低信道传输音频的听觉质量。针对此问题,提出一种结合密钥和随机标准正交基的音频伪装方法。首先对秘密音频和公开音频分段,利用密钥构造随机标准正交基;其次通过秘密音频小段在随机标准正交基上的投影来对秘密音频小段进行表达,从中选取包括均值幅值较大的前k个投影系数,并记录对应的索引位置;再次通过EMD-q密写方法嵌入到对应的公开音频小段中形成信道公开传输音频;最后通过信道公开传输音频提取的变换参数结合密钥重构秘密音频。实验表明,所提方法可充分利用随机标准正交基重构不同精度的秘密音频,且随着选取的幅值系数增多,恢复的秘密音频质量也越来越好,同时所述策略严格依赖于密钥,只有掌握正确密钥的用户才能进行高精度的重构。

The transformation model adopted in traditional tangram based audio camouflage method is the affine transformation model,where its transformation accuracy is low and doesn't meet the basic orthogonal relationship.It cannot be guaranteed the fitting accuracy between the secret audio and the public audio in affine transformation model.At the same time when the segment of transforming audio is constant value sequence,it must be added random disturbance to ensure the quality of audio recovery,which will reduce the audio's quality transmitted in channel.To solve these problems,an audio camouflage method combining key and random orthonormal basis is proposed.The secret audio and public audio are firstly segmented,and a key is used to construct the random orthonormal basis.Secondly,the secret audio segment is effectively expressed by the different projections of secret audio segment on the random orthonormal bases.The first k projection coefficients with larger amplitude included mean value are selected and their corresponding index positions are also recorded.Then the EMD-q steganography is employed to embed them into the corresponding public audio segment to form the public audio transmitted in the channel.Finally,the secret audio can be reconstructed by the transformation parameters extracted from the transmitted audio and the key.Experiment shows that the proposed method can make full use of the random orthonormal bases to reconstruct the secret audio with varying precision degrees,and the more larger amplitude coefficients are selected,the higher quality of the recovered secret audio will be.Simultaneously,the strategy described relies strictly on the key and only the user who masters the correct key can get the recovered secret audio in high quality.