US 12,093,314 B2
Accompaniment classification method and apparatus
Dong Xu, Guangdong (CN)
Assigned to Tencent Music Entertainment Technology (Shenzhen) Co., Ltd., Guangdong (CN)
Filed by Tencent Music Entertainment Technology (Shenzhen) Co., Ltd., Guangdong (CN)
Filed on May 19, 2022, as Appl. No. 17/748,494.
Application 17/748,494 is a continuation of application No. PCT/CN2020/128751, filed on Nov. 13, 2020.
Claims priority of application No. 201911155290.6 (CN), filed on Nov. 22, 2019.
Prior Publication US 2022/0277040 A1, Sep. 1, 2022
Int. Cl. G06F 16/65 (2019.01); G10H 1/36 (2006.01)
CPC G06F 16/65 (2019.01) [G10H 1/36 (2013.01); G10H 2210/005 (2013.01); G10H 2210/036 (2013.01); G10H 2250/311 (2013.01)] 17 Claims
OG exemplary drawing
 
1. An accompaniment classification method, comprising:
obtaining a first type of audio features of a target accompaniment, the first type of audio features comprising at least one kind of audio features, the first type of audio features comprising at least one kind of Mel spectral features, relative spectral transformation-perceptual linear prediction (RASTA-PLP) features, or perceptual linear prediction (PLP) coefficients;
performing data normalization on each kind of audio features in the first type of audio features of the target accompaniment to obtain a first feature-set of the target accompaniment, the first feature-set comprising at least one kind of audio features;
inputting the first feature-set into a first classification model for processing, the first classification model being a convolutional neural network model;
obtaining a first probability value output by the first classification model for the first feature-set;
determining an accompaniment category of the target accompaniment to be a first category of accompaniments when the first probability value is greater than a first classification threshold; and
determining the accompaniment category of the target accompaniment to be other categories of accompaniments when the first probability value is less than or equal to the first classification threshold,
wherein determining the accompaniment category of the target accompaniment to be other categories of accompaniments comprises:
obtaining a second type of audio features of the target accompaniment, the second type of audio features comprising the first type of audio features, and the second type of audio features further comprising at least one kind of: spectral entropy, first-order difference coefficients of the RASTA-PLP features, or second-order difference coefficients of the RASTA-PLP features;
determining a second probability value according to the second type of audio features of the target accompaniment; and
determining the accompaniment category of the target accompaniment to be a second category of accompaniments when the second probability value is greater than a second classification threshold.