HighLight text.png

Project active
Fall 2020 - Spring 2021

Problem: 

In the heat of the game, players cannot tell the difference between pocket and rim shots.

A modular Spikeball attachment that alerts the player of an illegal rim shot through visual and audio cues.

Solution:

Summary

final_logo_square.png

Spikeball is a fast, dynamic game in which two teams have up to three hits to bounce the ball off the net to the other team. A bounce off the net is legal, and play continues. However, a bounce off the rim is illegal and the other team gets a point. Shots that are right near the edge of the rim, called pocket shots, are legal, but during the heat of the game, it can be difficult to determine if the rim was hit. This issue is often a source of conflict during Spikeball games, resulting in arguments and loss of game time. HighLight seeks to fix this problem by serving as an unbiased referee to differentiate between pocket and rim shots.

HighLight is a modular attachment which straps around one of the legs on a Spikeball net. It utilizes machine learning and vibration patterns to detect the difference between rim and net shots. On a rim shot, it lights up and makes a noise to notify the players of the illegal hit. This allows players to focus on the game while HighLight provides a fair, unbiased referee to the game.

IMG_6125.HEIC

Current Progress

_DSC6988.JPG

We currently have a working prototype for HighLight with ~80% accuracy. It utilizes an Arduino Nano BLE 33 which gathers onboard accelerometer data and compares against a neural network trained using Tensorflow Lite. The unit is housed in a 3D printed PLA housing which straps around one of the legs on the Spikeball net. There is a 12 segment LED ring and 3.3V speaker in the housing which is used for the visual and auditory cues. The unit is battery-powered, so there are no cords to get in the way during gameplay.

Future Goals

We would like to improve the detection accuracy to improve the user experience. This could be achieved by utilizing a more secure attachment system, modifying the machine learning model, and gathering more training data. Another option we are considering is developing an integrated model where we install electronics into each leg of the Spikeball net. This could improve the accuracy due to more precise readings. We are also looking into new features such as different modes for different playing surfaces (sand, turf, or grass). We would also like to make the unit waterproof so that it can be used in all conditions.