Singing evaluation method and system

Abstract

The invention relates to the technical field of speech signal processing, and discloses a singing evaluation method and system. The method comprises the following steps of: receiving a continuous speech signal of singing of a user; determining the boundary location of each note in the continuous speech signal so as to segment each note; selecting the tone model corresponding to each segmented note, and obtaining the tone model through training in an off-line mode in advance; extracting the tonal character of the note; according to the tonal character of the note and the tone model corresponding to the segmented note, calculating the standard of the note; and according to the standard of the note, determining a singing evaluation score. According to the method and the system, the objectivity and accuracy of note matching can be increased, and the evaluation performance of the singing evaluation system is increased.

Description

technical field [0001] The invention relates to the technical field of voice signal processing, in particular to a singing evaluation method and system. Background technique [0002] In recent years, some karaoke products give score after the user sings, which improves the interactive entertainment. Traditional singing scoring systems usually use the method of comparing the difference between the pitch curve of the user's singing voice data and the preset standard pitch curve, by directly calculating the size of the error area between their fundamental frequency curves or using Dynamic Time Warping (Dynamic Time Warping, DTW for short) and other technologies align the fundamental frequency curves and calculate the difference between the fundamental frequency curves to evaluate the level of singing. The larger the curve difference value, the lower the user's singing data score; otherwise, the higher the user's singing data score. This method easily leads to the problem of u...

AI Technical Summary

Problems solved by technology

[0008] This method based on pitch curve fitting needs to estimate more parameter values ​​when calculating the true pitch of each note, and in the case of limited note duration (only 0.1 seconds), the amount of training data is relatively small, Therefore, it is easy to overfit and lead to insufficient robustness of parameter estimation, which affects the evaluation performance of the system

Images