【学术报告】5月15日: Zhaoxian Zhou报告会
1.题目：Human Activity Recognition and Sports Performance Evaluation
Abstract: Human activity recognition (HAR) has been one of the important fields of research in recent years. Activity recognition methods can be basically classified into three major types: radio-based, sensor-based; vision-based. Recently deep learning (DL) has been emerged as an approach for sensor-based human activity recognition but still lacks the significant increase in the accuracy for activity recognition than other machine learning approaches. The performance of DL methods depends largely on the feature extraction and classification techniques. Despite the vast development of HAR algorithms, the problem remains challenging for real time applications. Many applications require a multiple activity recognition system. For example, in automatic sports categorization, video streams alone are not efficient in recognizing athlete activities. Data from IMU sensors will help implementing faster and more accurate algorithms. Sensor-based algorithms should be incorporated with vision-based ones to improve the accuracy, efficient, and robustness. Another application is crowd control and public safety in large scale venues such as sports stadium and arenas. Multisource methods can accelerate the HAR process and thus provide more accurate and prompt decisions.
Zhaoxian Zhou, professor of the University of Southern Mississippi。
Zhaoxian Zhou received the B. Eng. from the University of Science and Technology of China in 1991; the M. Eng. from the National University of Singapore in 2000 and the PhD degree from the University of New Mexico in 2005. All degrees are in electrical engineering. From 1991 to 1997, he was an electrical engineer in China Research Institute of Radiowave Propagation. He joined the School of Computing at the University of Southern Mississippi in 2005. His current research interests include HAR, face recognition, and image processing; big data analytics, machine learning, computational sciences, and electrical engineering。