Research on virtual theater actor character performance based on machine learning

Theater, a timeless storytelling medium, is evolving through digital technology and virtual reality, enabling virtual theater to transform live performances into immersive, interactive, and boundary-breaking digital experiences. Research explores the use of machine learning (ML) techniques to enhance actor character performance in virtual theater settings. Virtual theater combines digital technology and performing arts to create immersive experiences where actors’ movements and expressions are translated into lifelike virtual characters. Research focuses on capturing and analyzing complex human behaviors, including gestures, facial expressions, and social engagement cues, using motion capture technology and advanced ML algorithms. Data is collected from actor performances using motion capture suits equipped with sensors that record full-body movements, gestures, and facial expressions. The data was preprocessed through low-pass filter, a Weiner filter and histogram equalization. The data undergoes feature extraction by utilizing Mel-Frequency Cepstral Coefficients (MFCC) and principal component analysis (PCA). Canonical Correlation Analysis (CCA) was used to fuse the features to prepare it for training models capable of replicating nuanced emotional and physical expressions in real time. An Artificial Gorilla Troops Optimizer-driven Bagging-enriched Extreme Gradient Boosting (AGT-Bagging XGBoost) algorithm is employed to interpret temporal dynamics and dynamically adapt virtual characters’ behaviors during live performances. The results indicating significant improvement, showed a Peak Accuracy of 99.13%, Error Rate of 0.55 and an emotional category accuracy of 98.21%, indicating that the ML-enhanced virtual characters achieve higher realism and expressiveness compared to traditional animation methods, thereby improving audience engagement and immersion.