Compressed Sensing And Its Applications In Audio Information Hiding

With the popularity of computers and the Internet, the digital signal has become the mainstream of signal processing. Sampling is the first and most crucial step in traditional analog/digital conversation:the sampling frequency must meet the requirements of the Shannon sampling theorem so that this sampling could fully recover the original signal. On the other hand, the rapid development of multimedia technology, the amount of multimedia data such as video,audio and text is increasing exponentially, which makes it difficult to meet the data storage by hardware devices. Compressed sensing theory is proposed to resolve data sample and compression. It breaks the Shannon sampling theorem and provides a new thought for signal processing. In the process of compressed sensing, the data is measured by the measurement matrix. The original signal can be restored correctly only when we know the measurement matrix. It also provides better protection for the security of the signal.From the initial discrete cosine transform, Fourier transform, wavelet transform to the latest adaptive dictionary learning, compressed sensing theory is developing constantly. The application of this theory is becoming more and more widely, such as image information security, communication coding, pattern recognition; face compression, astronomy and medical imaging and sensor data. This paper mainly studies information hiding in the speech signal based on compressed sensing. It divided into the following sections:(1) Introduce sparse representation method, selection of measurement matrix and reconstruction algorithms. Then, introduce the dictionary learning theory which applied to the sparse representation of signals.(2) Introduces the basic method of information hiding in the speech signal, the generation process of digital speech signals, the methods and steps of speech information hiding. And the feasibility that the compressed sensing theory could be applied to the speech signal processing.(3) Encrypted speech signal based on compressed sensing. Describes two methods of speech encryption based on compressed sensing: one is the traditional encrypting way of compressed sensing; another is redundant dictionary encryption by K-SVD dictionary learning algorithm.(4) Audio watermarking algorithm based on compressed sensing. The denoising method of compressed sensing is applied to the audio watermark embedding and extraction.To a certain extent, this method could improve the robustness of the watermark.