Audio Steganography using Modified Sparse Representation Based Wavelet Transform

A technique There are three main problems in designing speech steganography technique: un detectability, imperceptibility, and capacity. In the process Un detectability the stego signal, containing the secret data, is in differentiable from the original speech signal neither by means of human auditory system (HAS) nor steganalysis techniques. In the process of Imperceptibility the hidden data is inserted which makes no audible distortion in the original speech signal, whereas the capacity of the cover signal gives the comparative amount of hidden data which is to be inserted into the cover speech signal. The goal of this work is presenting a speech steganography technique which is less detectable, more secure, against steganalyzers than existing techniques. The novelty of this work comes by the employment of sparse representation of wavelet coefficients to embed secret data. Sparse representation technique makes it possible to embed the secret data in the higher semantic levels of the speech signal which leads to a more secure speech steganography technique. Sparse representation can also improve the quality of previous LSB coding technique.