Personalized song recommendation method based on voice timbre

Abstract

The invention provides a personalized song recommendation method based on voice timbre, relates to a network platform singing song recommendation method, and solves the problem that a collaborative filtering method is not suitable in the case and a speaker recognition correlation method cannot be used directly, and thus large influence is caused on the recommendation result. The method comprises the following steps: adopting accompaniment and voice separation; 2) extracting acoustic features MFCC and LPCC capable of representing user timbre features from the voice data; 3) recommending personalized required songs according to the acoustic features extracted in the step 2); and 4) carrying out expandability analysis. The method improves user experience in a network singing system and has a very good application prospect.

Description

technical field [0001] The invention relates to a singing-demand recommendation method on a network platform, in particular to a personalized singing-demand recommendation method based on human voice. Background technique [0002] At present, with the vigorous development of online music, people's requirements for music services are getting higher and higher. Major music service websites have successively launched personalized music recommendation functions, that is, through historical behaviors such as user access behaviors and collection records. Analyze and mine users' interests and hobbies, and recommend music that matches their appreciation taste. [0003] Online singing is a music service product that has developed rapidly in recent years. It moves the traditional KTV singing function to the Internet and provides a virtual singing platform for singing lovers through the Internet. Since most of the users of online KTV are amateur users, they do not have rich music know...

AI Technical Summary

Problems solved by technology

[0005] In order to solve the problem that the most widely used collaborative filtering method in traditional music recommendation is not applicable here, the present invention recommends songs of singers with similar timbres to the user based on the timbre characteristics of the user’s own timbre. , so it is different from speaker recognition, and the relevant methods of speaker recognition cannot be directly used; the recommendation can only be based on the vocal information of the singer himself, and the accompaniment and harmony will affect the recommendation results, providing a The specific technical implementation plan of the personalized on-demand singing recommendation method based on the human voice is as follows:

Images