Research On Digital Right Retrieval Based On ANN

With the development of science and technology, digital content has been widely used in many fields, such as education, medical and entertainment, etc., it has to involve the issue of data content copyright management. Because of the digital multimedia is easy to edit in digital copyright retrieval system based on content, reasonable editor or illegal tampering with the copyright of multimedia leads to extract the identifiers or content feature codes and the original records stored data will has certain deviation, this needs to look for in the copyright characteristic database with a given digital content copyright information of "similarity", the search process is called "similarity retrieval". How to rapidly retrieved in the large database of copyright content based similarity retrieval is becoming more and more important. The traditional nearest neighbor queries are affected by the "dimension disaster", along with the increase of data dimension, the performance of traditional index structure fell sharply.According to above problem, considering the approximate nearest neighbor search algorithms (ANN) to solve the problem of searched in high-dimensional space. By using the method of approximate nearest neighbor can quickly get the point set similar with the requirements of retrieval, a kind of important method to solve the problem of ANN is the Locality Sensitive Hashing (LSH) algorithm. LSH algorithms is based on a hash index on the basis of ANN search algorithm, it does not rely on digital content, the characteristic dimension of the other based on Tree data structure, such as R-Tree, KD-Tree, SR-Tree, it is better to overcome the dimension disaster and reduce the time complexity of nearest neighbor query to sub-linear, so LSH can effectively solve the problem of ANN search in high-dimensional feature vector.The main work of this paper is:study and learn the basic principle and method of LSH algorithms, according to the computation of faster hash function and skip the repeat points, the LSH algorithms are optimized, and then through the experimental data of LSH algorithms and improved the traditional query performance index method has carried on the detailed comparison, the experimental results show that the improved LSH algorithms under the condition of without reducing accuracy, spend less time and improve the efficiency of the query. After analysis and get the following conclusion, the improved LSH algorithms used in the digital copyright retrieval based on content, its performance is superior to the traditional method.