Research On Digital Audio Watermarking Algorithms

As a research hotspot in the field of information security, information hiding technique has developed rapidly in recent years, to which many scholars and experts have already paid much attention. It includes two aspects: covered communication and digital watermarking. Nowadays, network makes communication more convenient, however, traditional encryption can't meet the need of information security. The easily-reproduced and easily-transmitted characteristics of digital multimedia make for flexibility in accessing information, however, it has also led to large-scale unauthorized copying.The thesis is on digital audio information hiding, it contains four parts:1.It reviews the evolution of information hiding, describes some basic items and main framework of information hiding and digital audio watermarking, focuses on the properties of audio watermarking, analyzes the proposed audio watermarking algorithms in the wavelet domain, and points out some promising directions in this research field.2.A new robust audio watermarking algorithm in DWT(Discrete Wavelet Domain) is proposed. Chaotic sequence has the properties of the initial value sensitivity, Gaussian Noise resembling and low correlation, so we use chaotic sequence as synchronization code embedded into audio together with the watermark. Experimental results show that under various attacks, the synchronization code can be detected. In the course of the preprocessing of watermark, another chaotic sequence is generated to modulate the original watermark in order to enhance the security degree of the watermark. Meanwhile, the algorithm applies error correction coding and image scrambling to the watermark, then the preprocessed watermark is embedded to the low frequency coefficients in DWT. When extracting watermark data, error correction mode is adopted. The algorithm realizes blind extraction.3.The above algorithm is improved to meet the covered communication's requirements of secrecy and abundant embedding. It embeds large numbers of data both in time domain and wavelet domain. An improved LSB embedding method is used in time domain, in which the embedding location is randomly selected by a chaotic sequence to keep secret. The experiments show the data can't be discovered and stand against common signal disturbing.4.We implement the above algorithms in software, an audio watermarking system, in which multiform watermarks and several kinds of the embedding algorithms are capable of being selected flexibly. The system enables the evaluating of watermarks. The secret data may be embedded in time domain or wavelet domain or both. Consequently the system can be applied widely in many areas.