Dance Action Recognition Model Using Deep Learning Network in Streaming Media Environment.

Area of Science:

• Computer Vision
• Artificial Intelligence
• Machine Learning

Background:

• Dance movement recognition is a key video technology for intelligent applications.
• Effective feature extraction is crucial for accurate dance movement recognition.
• Deep learning offers advanced methods for video feature extraction.

Purpose of the Study:

• To investigate dance movement recognition using deep learning-based convolution neural networks.
• To evaluate the effectiveness of convolution neural networks in extracting dance movement features.
• To improve the accuracy of dance movement recognition systems.

Main Methods:

• Utilized deep learning networks, specifically convolution neural networks (CNNs).
• Implemented CNNs for feature extraction from dance videos.
• Integrated manually extracted time-domain optical flow information with CNN models (InceptionV3 and 3D-CNN).

Main Results:

• The proposed CNN method demonstrated viability for dance movement recognition.
• Accuracy increased by 30.65% (InceptionV3) and 19.49% (3D-CNN) with added optical flow information.
• The CNN approach proved more effective for identifying dance movements compared to baseline methods.

Conclusions:

• The developed convolution neural network method is effective for dance movement recognition.
• Improved recognition systems have significant potential in dance instruction and practice.
• This research shows promising application potential for advanced dance recognition technology.