Abstract

The research community has become more interested in human activity recognition due to improvements in technology and machine learning algorithms. In particular, when we discuss the automated assessment of athletic talents, which has been the most active study topic over the last decade. The highly competitive nature of sports necessitates collecting accurate data on an athlete’s performance to evaluate their activities while competing accurately. In this article, we present a method for identifying seven typical styles of Mae Mai Muay Thai (MMMT) using continuously collected boxing sequences as the data source. Long-term Recurrent Convolution Networks (LRCNs) is used to handle MMMT recognition. Additionally, we combined Long Short Term Memory (LSTM) and Convolutional Neural Networks (CNNs) classifiers for experimental testing. According to experimental testing, our training strategy outperformed the use of both CNNs and LSTM classifiers. Experiments were conducted utilizing the MMMT dataset with four professional boxers as participants. The LRCNs classifiers were able to achieve a 99 percent accuracy rate that proves the LRCNs algorithm is appropriate for assessing the boxer’s abilities while competing. Furthermore, we will determine the overall usefulness of the model by using a confusion matrix in our analysis. Additional performance metrics included in the investigation were accuracy, precision, recall, and the F1-score.