A Humming Retrieval System for Massive Music Data

Abstract

The invention discloses a humming retrieval system for massive music data, comprising a template feature extraction module for extracting melody features in database template files; a humming feature extraction module for extracting melody features of humming files; The module is used to retrieve the template pitch vector; the effective alignment position detection module is used to obtain the effective alignment position; the matching module is used to match the query pitch curve and the template pitch curve to obtain a list of candidate songs. On the one hand, the present invention greatly reduces the number of pitch vectors by detecting feature points in the template feature extraction module and humming feature extraction module; Reduce the number of matches to a great extent and improve the retrieval speed; on the other hand, in the matching module, the improved DTW algorithm is used to solve the problem of accurate local humming by users and improve the retrieval accuracy.

Description

technical field [0001] The invention relates to the fields of digital signal processing technology, basic music theory and machine learning technology, in particular to a humming retrieval system for massive music data. Background technique [0002] With the rapid development of high-capacity digital storage and digital music technology, content-based multimedia search technology has gradually become a topic of concern to researchers. Under the background of the rapid popularization of the Internet and the development of storage devices, the digital music resources available to users have grown rapidly. How to retrieve and obtain the music required by users from the massive music database has become an urgent problem to be solved, which has also created a humming retrieval system. appearance. [0003] The traditional humming retrieval system for massive music data has the following shortcomings: [0004] 1. There are redundant template pitch vectors and query pitch vectors...

AI Technical Summary

Problems solved by technology

[0004] 1. There are redundant template pitch vectors and query pitch vectors, resulting in an increase in the number of matches;

[0005] 2. The alignment position of the pitch vector is not filtered, resulting in many invalid matches;

[0006] 3. Matching is performed under the condition that the two ends of the query pitch curve are not really aligned in the template pitch curve, resulting in large errors at both ends and low matching accuracy;

[0007] 4. Unable to deal with the problem of partial accuracy of humming

Images