as
X
Tongue trackpad writing
tongue trackpad device
The "Tongue Trackpad Writing" project aimed to develop a software application that interfaces with a specialized touchpad designed to be controlled with a tongue. the primary objective of this project was to create a system that interprets tongue movements on the touchpad and predicts and outputs the letters of the alphabet being written by the user.
tongue trackpad writing user interface
html and javascript was used to create the user interface and read data
from the device through bluetooth. python was used to implement the letter recognition software. google
colab and ngrox was used to host the model online for testing purposes.
Initially, we used the NIST handwritten dataset but faced issues due to the distinct appearance of tongue-written letters. Collaborating with fellow researchers, we gathered tongue writing samples to enhance the dataset, resulting in improved recognition rates. To further enhance our system, we implemented an auto-correction feature that deciphered guessed sentences and achieved impressive accuracy in multiple tests.
1-sentence tongue writing, guessed text, and autocorrected text
original text was: THE FIVE BOXING WIZARDS JUMP QUICKLY
only three out of 31 characters guessed incorrectly
original text was: THE FIVE BOXING WIZARDS JUMP QUICKLY
only three out of 31 characters guessed incorrectly
