Classification of K-Pop Dance Movements Based on Skeleton Information Obtained by a Kinect Sensor

Area of Science:

• Computer Science
• Artificial Intelligence
• Robotics

Background:

• K-pop dance classification is challenging due to complex and dynamic movements.
• Existing methods often struggle with the high dimensionality and subtle variations in dance data.
• Automated dance analysis requires robust feature extraction and efficient classification techniques.

Purpose of the Study:

• To develop an automated method for classifying K-pop dances using skeletal motion data.
• To create a comprehensive K-pop dance database for research and development.
• To evaluate the performance of the proposed classification method against conventional approaches.

Main Methods:

• A K-pop dance database was constructed with 800 movement data points from 200 dance types.
• Six core angles representing motion features were extracted from skeletal joint data.
• Dimensionality reduction was achieved using a combination of Principal Component Analysis and Fisher's Linear Discriminant Analysis (fisherdance).
• A Rectified Linear Unit (ReLU)-based Extreme Learning Machine Classifier (ELMC) was employed for classification.

Main Results:

• The proposed method achieved superior classification performance compared to K-Nearest Neighbor (KNN), Support Vector Machine (SVM), and standalone Extreme Learning Machine (ELM).
• The fisherdance technique effectively reduced data dimensionality while preserving crucial motion information.
• The ReLU-based ELMC demonstrated rapid processing times without requiring weight learning.

Conclusions:

• The proposed skeletal motion-based K-pop dance classification method is effective and efficient.
• The developed fisherdance and ELMC approach offers a promising solution for automated dance analysis.
• This research contributes to the field of human motion analysis and pattern recognition.