Multiple Audio Database Watermarking Algorithm Based On Blind Source Separation

With the development of computer resource sharing and networking technology, there are more and more applications of database in network. Database stores a large amount of useful information and the information has great social value and economic value. If the information is stolen by the wrong crowd, the owner of the database will suffer incalculable losses. Under such circumstances, how to protect copyright of database becomes a serious problem. By embedding watermark with copyright information in database, database watermarking can link databases with their owners and protect the copyright of databases.Most of the existing database watermarking algorithms embed a single watermark in the database, but this can not meet people's needs. In many cases, what is needed is a multiple watermarking or a multi-functional watermarking. This paper reviews the development of watermarking relational database, and outlines basic theory of digital watermarking and watermarking relational database, including the concept of digital watermarking, general models, the characters, attacks and the key issues solved by database watermarking. To solve the inadequacy of existing algorithms, this paper analyzes the existing database watermarking algorithm and proposes two kinds of watermarking algorithm:(1) A multiple audio watermarking for database based on blind source separation. As large databases now tend to be completed by a multi-party co-operation, it is a necessity to protect the interests of multiple copyright owners. In this algorithm, the watermarking, a mixture of the speech signals of multiple copyright owners is inserted into database. After multiple watermarking is extracted, the speech signals of copyright owners will be separated by blind source separation algorithm based on maximum signal noise ratio. Then the speeches are matched by speech recognition technology based on DTW algorithm in order to prove that speeches are owned by multiple copyright owners. In this way, the interests of copyright owners are effectively protected.(2) Based on the above-mentioned algorithm, this paper tries to apply a multi-functional watermarking algorithm for database to the demanding occasions. Multiple audio watermarking with the information of copyright owner is embedded in the numerical attributes of database to protect the interests of multiple copyright owners and enhance the robustness of database watermarking algorithm; fragile watermarking is embedded in the non-numerical attributes of database. When the database is attacked, the watermarking can be used to authenticate the integrity of the database and locate the tampering region.This paper focuses on a multiple audio watermarking for database which is based on blind source separation and tries a multi-functional database watermarking algorithm. Both watermarking algorithms include three processes:watermarking generation, watermarking embedment and watermarking extraction. Simulation experiments are carried out to verify the capacity of watermarking algorithm in resisting all kinds of malicious attacks. Theoretical analysis and experiments approve that the proposed watermarking schemes in this paper are effective and have the practical and theoretical significance for copyright protection of databases.