Action Sports Learning Based on Expert Instruction Revised by Wearable Sensor Data
2018
An instruction of expert is one of the important factors for action sports learning of a student and sports education.Video movie recording is also an effective tool to recognize skill level visually.However, the student might misunderstand the instruction because of ambiguity of natural language.Moreover, it is quite difficult to recognize an extremely quick and small action with the video movie.As a practical solution to overcome these problems, this paper reports an example of action sports learning with wearable sensors and time series data analysis.The sensor is composed of three-axis accelerometers, three-axis gyroscopes, three-axis digital compasses, and GPS.To categorize actions based on multiple time series data collected from the sensors, we also developed a classification method of time series data with local cross-correlation function.We apply our method to analyze BMX flatland riding.According to the result, the method can successfully categorize ten types of tricks into appropriate sub categories.We also depict motion timing and sequence among all body parts of an athlete to understand the tricks.Based on the sensor data, the expert can revise the instruction to make the student learn proper way of action sports performance.