Research on Classification and Recognition of Badminton Action Based on Spectrum Analysis

EasyChair Preprint 5349

Abstract

Badminton, one of the most favorable sports of all. As of 2020, there are over three million population in Taiwan used to play badminton according to the Sports Administration. If a guideline of long ball, smash ball and drop ball are developed, it would be instructive for beginners to improve their badminton skills. Since smart watches are gradually integrated into daily life, so this paper uses the three-axis acceleration sensor of smart watches to collect the data of user's swing movements to distinguish the three types of swing movements of badminton. It was not an easy way to distinguish the type and the level of the shot if the data were just analyzed in the time domain. Therfore, we use  the characteristics of the spectrum to differentiate after transferring the data to the spectrum. Then using a visual tool to analyze the differences between the national teams and the badminton enthusiasts, and using decision tree to create the classification. The results show that after the data is visually analyzed, important attributes are captured for classification, and the accuracy is improved.

Keyphrases: Decision Tree, acceleration sensor, smart watch

@booklet{EasyChair:5349,
  author    = {Mei Hui Hsu and Jui-Chung Hung},
  title     = {Research on Classification and Recognition of Badminton Action Based on Spectrum Analysis},
  howpublished = {EasyChair Preprint 5349},
  year      = {EasyChair, 2021}}