Multimedia resource acquisition method, apparatus and system

Abstract

The invention discloses a multimedia resource obtaining method which comprises the steps: a multimedia resource saved in the form of vector points after hash process is stored by a first server uniformly; the first server determines a query point specified by a user and an identifier of a hashing bucket that the query point belongs to according to the request message transmitted by a client, and transmits a notification message carrying the identifier to a second server; and when determining the local existence of the hash bucket saved corresponding to the identifier according to the identifier carried in the notification message, the second server obtains at least one vector point nearest to the query point in the hash bucket and sends the at least one vector points to the client. In this way, operational pressure on the first server is released to a degree, thereby quickening the responding speed of a whole system in providing multimedia resource query service, and further improving user experience. The invention also discloses a communication device and a communication system simultaneously.

Description

technical field [0001] The present invention relates to the field of computers, in particular to a method, device and system for acquiring multimedia resources. Background technique [0002] With the popularity of multimedia resources (including audio resources, video resources, etc.) in Internet applications, how to quickly and accurately find the required multimedia resources in the massive network resources has become an urgent problem to be solved. [0003] At present, multimedia resources on the Internet are generally stored in a general method. The general method is: divide a multimedia resource into several key frames, and for each key frame, use the CBIR (Content Based Image Retrieval) method Extract several pieces of key information from key frames to obtain a vector point represented by a set of numbers, and a multimedia resource can be represented by a certain number of vector points obtained from several key frames through CBIR processing. Next, save the vector ...

AI Technical Summary

Problems solved by technology

[0006] At present, a very large number of high-dimensional vector points are stored in the database on the central server, and the number of vector points will change frequently with the update of multimedia resources (including addition / deletion / modification). Obviously, when the database When the number of vector points stored in each hash bucket is too large, the central server, at the request of the user, linearly queries the vector point closest to the query point proposed by the user in each hash bucket, which will consume a lot of CPU resources and seriously It affects the response speed of the entire system, which seriously reduces the user experience

Images