Martial Arts Movement and Posture Recognition using Quantum Multi-Verse Optimization based Long Short-Term Memory

In recent years, the exceptional growth of sports and globalization of martial arts events has resulted in a constant improvement in competition regulations. However, it is critical to detect and correct technical errors in martial art. As a result, a visual image-based technique is necessary to consistently and effectively detect martial arts movements. As a result, this paper proposed utilizing Quantum Multi-Verse Optimization-based Long Short-Term Memory (QMVO-LSTM) to distinguish human martial arts motions and postures using image recognition. Then the edge parts of human body curve are detected after the bulk of the background noise has been removed using the morphological gradient approach. Next, dynamic temporal warping is performed to compare the classifier and joint change sequences across diverse martial arts activities. Finally, the suggested QMVO-LSTM is used to image recognition in order to evaluate and categorize human martial arts abilities. According to the results, it illustrates that the suggested QMVO-LSTM model has achieved better in terms Accuracy of 99.88% and recall of 99.86% which is comparatively higher when compared to existing models such as Attention enhanced Spatial Temporal Graph Convolutional LSTM (ASTGC-LSTM), Two-Dimensional Principal Component Analysis (2D-PCA) and Convolutional Neural Network (CNN).