Identifying and Animating Movement of Zeibekiko Sequences by Spatial Temporal Graph Convolutional Network with Multi Attention Modules

Folk dances, integral components of intangible cultural heritage (ICH), are both fleeting and fragile. However, with the rapid advancement of computer vision techniques, there arises an opportunity to document and safeguard these cultural expressions for future generations. This study aims to identify the distinctive dance sequences and characteristics of Zeibekiko, a popular Greek folk solo dance found in variations across Greece, Cyprus, and the Aegean region of Minor Asia, and translate them into a virtual 3D environment. Utilizing a state-of-the-art optical motion capture system featuring active markers (the PhaseSpace X2E system), precise recordings of the Zeibekiko dance are achieved. The three-dimensional spatial data derived from the dancer's movements serves as the foundation for classification, accomplished through a Spatial Temporal Graph Convolutional Network with Multi Attention Modules (ST-GCN-MAM). This innovative architecture strategically employs attention modules to extract key features of the dance from primary areas of the upper and lower parts of human body. With high level accuracy, the proposed tool accurately detected and recognized Zeibekiko sequences. Ensuring the precise alignment of captured points with corresponding bones or anatomical features in the 3D dancer model is essential for seamless and authentic animations. Advanced visualization and animation techniques are then employed to translate these points into smooth, realistic character movements, preserving their inherent dynamics and expressions. As a result, a faithful virtual rendition of the dance is achieved, capturing its authenticity and beauty. Such a solution holds potential applications in gaming, video production, or virtual museum exhibits dedicated to showcasing folk dances.

---

Opracowanie rekordu ze środków MNiSW, umowa nr POPUL/SP/0154/2024/02 w ramach programu "Społeczna odpowiedzialność nauki II" - moduł: Popularyzacja nauki (2025).