[Retracted] Application of Machine Learning to Badminton Action Decomposition Teaching

Table 2

Simulation experiment.

Experiment names	Experiment codes	Experimental classification	Description
Comparison of three different segmentation methods	A	Sliding window segmentation in the hitting time	50 sample points before and after hitting the ball as action characteristics
	B	Sliding window segmentation based on the maximum	Sliding windows with a width of 100, the maximum, and the reset
	C	Window segmentation based on events	The starting and ending time of the player’s action as those of the experiment
Training and recognition models of different algorithms	D	SVM	Identifying ten hitting actions
	E	Hidden Markov model (HMM) [29]	Comparison of the posterior probability of the model after training
	F	Combination of Haar-like, AdaBoost, and HMM
The system recognition rate of images	G	One player	Training and test data of one player
	H	Different players	Training and test data of many different players