Large-scale humming melody matching system based on data level paralleling and graphic processing unit (GPU) acceleration

Abstract

The invention provides a large-scale humming melody matching system based on data level paralleling and graphic processing unit (GPU) acceleration and aims at solving the problem that the existing practical large-scale humming retrieval system is lacked and is low in performance and the like. The system mainly comprises a humming melody extraction module, a humming retrieval melody base building module and a humming melody matching module based on data level paralleling. Due to the fact that the system faces a newly-developing application of large-scale humming retrieval, the new technologies of data level paralleling, and GPU acceleration and the like are completely utilized, forceful support is provided for performance improvement of the large-scale humming retrieval system, and the system has wide application prospects.

Description

technical field [0001] The invention discloses a large-scale humming melody matching system based on data-level parallel GPU acceleration, and relates to humming retrieval and parallel acceleration technology. Through the method and device, users can accurately and efficiently retrieve desired music. It belongs to the field of computer technology. Background technique [0002] With the increase of the amount of information and the improvement of users' search requirements, the search based on media content has attracted more and more attention. The needs of music retrieval are diverse, and with the advancement of technology and the development of the times, this diversity has become more and more prominent. In terms of retrieval methods, music retrieval can be performed by submitting query requests in various ways such as humming, beating, performance input, sheet music entry, and audio samples. Among them, performance input retrieval and score entry retrieval can be easil...

AI Technical Summary

Problems solved by technology

[0004] The main problem of content-based music retrieval at present is the lack of user experience. Because the retrieval algorithms used in retrieval are generally more complex, the time required for retrieval is relatively long, and the waiting time for users is correspondingly long.

However, some existing patents (such as No. 200410049328.9, No. 200610065752.1, No. 200610122430.6, No. 200910301120.4, No. 201010144497.6, etc.) only involve a certain key technology of the humming retrieval system, and some (such as No. 200710301762.5) Although the humming retrieval system is considered, Marsyas technology and GPU parallel acceleration methods are not involved in the system

In general, there is a relative lack of existing practical large-scale humming retrieval systems and large-scale humming retrieval systems, especially the problem of low melody matching performance

Images