Secured Speech Watermarking with DCT Compression and Chaotic Embedding Using DWT and SVD

In this paper, a secured watermarking algorithm based on chaotic embedding of speech signal in discrete wavelet transform (DWT) domain of cover audio is proposed. The speech signal to be embedded is compressed using discrete cosine transform (DCT) by finding the suitable number of DCT coefficients such that the perceptual quality of decompressed signal is preserved. The chaoticmapis used to select the cover audio frames randomly instead of performing sequential embedding. The cover audio is decomposed usingDWTfollowed by singular value decomposition (SVD), and theDCT coefficients of the speech signal are embedded in the singular matrix of the cover audio. The proposedwatermarking algorithm achieves good imperceptibility with an average SNR and ODG of 46 dB and -1.07, respectively. The proposed algorithm can resist to various signal processing attacks such as noise addition, low-pass filtering, requantization, resampling, amplitude scaling, and MP3 compression. Experimental results show that the secret speech is reconstructed with an average perceptual evaluation of speech quality (PESQ) score of 4.26 under no attack condition, and above 3.0 under various signal processing attacks. Further, the correlation between original and reconstructed secret speech signal is close to unity. In addition, the loss in the generality of the information of the reconstructed speech signal is tested and is found minimum even the watermarked audio is subjected to various signal processing attacks. The proposed algorithm is also tested for false positive test to ensure the security of watermarking algorithm.