Dance Performance Recognition and Analysis Based on Pulse Coupled Neural Network

Dance is universally recognized expressive movement and it requires ongoing training and accurate evaluations to improve the expression and performance. The existing method struggled with capturing spatial data due to its long-term dependencies. Therefore, this research proposes a Pulse Coupled Neural Network (PCNN) for dance performance recognition. The PCNN is robust to noise which ensures the network to concentrate on essential features of dance moments and enhancing the recognition accuracy. It has capability to mimic the visual cortex behavior which is helpful to recognize image structures. The Histogram of Oriented Gradients (HOG) is used to extract shape and gradient features which is useful for shape pattern recognition. The Indian dance form recognition dataset is preprocessed by data augmentation which enhances the data size and prevent overfitting issues by introducing changes such as flips, rotations and translations. The PCNN achieved 99.95%, 99.66%, 99.52%, 99.93% and 98.16% of accuracy, f1-measure, precision, recall and AUC metrics compared to Fusion Neural Network (FUSNN).