Vibrato identification method and device

Abstract

The invention provides a vibrato identification method and device. The method comprises the following steps of obtaining to-be-detected audio data; intercepting a valid audio detection segment from the to-be-detected audio data; when the duration of the valid audio detection segment is longer than the minimum vibrato duration requirement, calculating the tones of all frames of the valid audio detection segment to obtain a tone sequence; grouping the tone sequence, and calculating the tone standard deviation and the target period of each group; deleting the group of which the tone standard deviation is greater than a preset standard deviation threshold value and the target period is greater than a preset vibrato period minimum value or less than a preset vibrato period maximum value; calculating the ratio of remaining groups in the total number of groups before deletion; and if the ratio is greater than a preset parameter, determining as a vibrato signal. By means of the scheme, the problem that the existing vibrato identification accuracy is not high is solved, and not only the vibrato identification accuracy is improved, but also the vibrato position and the vibrato duration can be accurately located.

Description

technical field [0001] The invention relates to the field of speech processing, in particular to a method and device for vibrato recognition. Background technique [0002] When extracting the pitch of the singing voice, we found that the pitch of a certain character is not a fixed value when singing, but fluctuates around a certain pitch value over time. This is due to the vibrato that the singer is infected by the emotion of the song and the singing skills during the singing process. The recognition of vibrato can be applied to the scoring benchmark for vibrato in singing scoring, enriching the singing scoring function. [0003] Existing tremolo extraction method is mainly by intercepting the sound segment to be detected, cutting off the head and tail fixed length of the sound segment, dividing the intermediate data into frames, and calculating the tone of each frame, and the obtained tone sequences are divided into different cycle lengths. Divide it into several segments...

AI Technical Summary

Problems solved by technology

[0005] The present invention provides a method and device for vibrato recognition, which effectively solves the problem of low accuracy of existing vibrato recognition, not only improves the accuracy of vibrato recognition, but also accurately locates the position and duration of vibrato

Images