Optical image embedding in speech signals with sensitivity analysis

This paper is mainly concerned with the embedding of optical images in auxiliary media. Optical images may contain sensitive information. They are embedded in cover media such as speech signals. This process is regarded as a type of watermarking. The Singular Value Decomposition (SVD) of 2D matrices generated from the cover media is used for watermark embedding. It is well-known that the Singular Values (SVs) of 2D matrices have low sensitivity to variations in the cover signal represented as noise or enhancement through processing algorithms. Noise affects the watermarked speech signal and affects the extraction of the watermark. Different enhancement algorithms are considered and compared for testing of the proposed scheme. It is clear from the obtained results that the proposed scheme is highly efficient for optical image hiding, even with signal processing techniques applied to cover signals. Simulation experiments indicate the effect of the presence of noise on the watermark extraction and also the effect of applying speech enhancement on the watermark extraction. The correlation coefficient (C-r) between the embedded and extracted watermarks is used to indicate the performance of different enhancement methods. The adaptive Wiener filter leads to the highest C-r, which equals 0.7491. Signal-to-Noise Ratio (SNR) is used to evaluate the speech enhancement performance. The SNR reaches the highest value equal to 12.0481 dB with adaptive Wiener filter.