The “Magic Glove” serves as the final project for our digital circuit laboratory.
Our goal is to design and create a glove equipped with multiple sensors capable of recognizing sign language in real-time.
To achieve this, we utilize an FPGA to implement a fast CNN (Convolutional Neural Network) model for sign language recognition. To improve recognition accuracy, we incorporate the Dynamic Warping algorithm and a built-in dictionary. The glove dynamically corrects recognized words that appear nonsensical within the vocabulary. We then employ the Viterbi algorithm to search the dictionary and identify the closest word to the original outcome.
In the hardware aspect, we integrate bending sensors on each finger and a gyro on the back of the hand. These sensors, along with an Arduino and Bluetooth module, are combined into the glove. When the glove is worn, sign language gestures are converted into real-time text through the “Magic Glove,” hence the name.
The communication between the Arduino and the computer occurs through Bluetooth. The computer relays this signal to the FPGA, which processes the data and displays the result on the screen with minimal delay.
Throughout this project, I gained valuable experience in Verilog programming and familiarized myself with the basics of FPGA usage. This “Magic Glove” project has been a fascinating exploration into the convergence of hardware and algorithms for a meaningful application in sign language recognition.