In order to achieve effective classification of dance movement in video standardization,a dance classification method based on Kinect sensor to obtain human skeleton motion data is proposed.Cascade vectors are generated from the six core angles of dance movements in each frame of the video.Principal Component Analysis(PCA)is used to project high-dimensional vectors into low dimensional space,and Linear Discriminant Analysis(LDA)is combined to obtain feature vectors with recognition ability.An Ex-treme Learning Machine Classifier(ELMC)based on the Rectified Linear Unit(ReLU)is designed.The ReLU function is used to solve the gradient disappearance problem in the neural network design,and the fast classification processing of dance movements is realized without weight learning.The experiment results show that compared with the traditional methods,the proposed method has better classification results,and the maximum classification rate can reach 96.5％.
关键词
主成分分析/线性判别分析/极限学习机/舞蹈动作/骨骼关节数据
Key words
principal component analysis/linear discriminant analysis/extreme learning machine/dance movements/bone joint data
维普
万方数据