This electronic paddle improves table tennis training

Table tennis, AKA ping pong, is a serious sport and like all serious sports in the modern age, its athletes can benefit from data-driven training. Without a coach watching, a table tennis player likely won’t even be able to keep track of the number of strokes made during a training session. To make that training more productive, Samuel Alexander built this table tennis electronic paddle that analyzes strokes.

Alexander put serious effort into making the smart paddle useful for real players and real training, so it has proper balance and rebound. But embedded within the handle’s grip is an Arduino Nano 33 BLE Sense Rev2 board that monitors the paddle’s movement. It uses that information to detect different kinds of strokes, keeping a tally of each that the player can then review on the OLED screen. 

The paddle tracks movement through the Arduino’s built-in LSM9DS1 IMU. But the data provided by that sensor is just an array of vectors and none of them correlates directly to a stroke. To turn the raw data into stroke tallies, Alexander turned to Edge Impulse to train a machine learning model based on real-world examples of each stroke type.

It can now reliably classify normal motion (not a stroke), backhand drives, backhand smashes, forehand drives, forehand smashes, and forehand loops. The count shows up on the grip’s OLED screen, but Alexander also created a web interface that users can visit with their smartphones to see more details, such as the session time and timestamps of strokes. 

Alexander found that this performs well, with an ultimate test accuracy of 88.7%.