Hybrid Activity Recognition for Ballroom Dance Exercise using Video and Wearable Sensor
2019 Joint 8th International Conference on Informatics, Electronics & Vision (ICIEV) and 2019 3rd International Conference on Imaging, Vision & Pattern Recognition (icIVPR)(2019)
摘要
In this paper, we propose a hybrid activity recognition method for ballroom dance exercise using video and wearable sensor. The purpose of our research is to design a mechanism to support ballroom dance exercise, and this paper reports the first part to design a mechanism is to support ballroom dance exercise, and this paper reports the first outcome to achieve the purpose - recognizing ballroom dance exercise. There are two conceivable ways to recognize dance exercise: videos and wearable sensor. However, each of them has its disadvantages. Using video is a good way to recognize the movement of the body. However, it cannot provide us accurate timing or strength of foot actions because the number of their flames per seconds is too small to recognize the fast movements of dancers. On the other hand, while a wearable sensor is good at recognizing foot timing and strength, it is not good at recognizing the movement of the whole body. Therefore we propose a hybrid recognition method utilizing the merits of both video and wearable sensor. This paper focuses to recognize four different types of steps in Latin American, a kind of ballroom dance. For each step, we record wearable sensing data and videos. As a result, it is found that the accuracy of step recognition is improved by adding wearable sensing data to video data shot from two different angles.
更多查看译文
关键词
Wearable sensors,Video signal processing,Machine learning,Activity Recognition
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络