people.ece.cornell.edu/land/co...
We designed and built a glove to be worn on the right hand that uses a Machine Learning (ML) algorithm to translate sign language into spoken English. Every person's hand is a unique size and shape, and we aimed to create a device that could provide reliable translations regardless of those differences. Our device uses five Spectra Symbol Flex-Sensors that we use to quantify how much each finger is bent, and the MPU-6050 (a three-axis accelerometer and gyroscope) is able to detect the orientation and rotational movement of the hand. These sensors are read, averaged, and arranged into packets using an ATmega1284p microcontroller. These packets are then sent serially to a user's PC to be run in conjunction with a Python script. The user creates data sets of information from the glove for each gesture that should eventually be translated, and the algorithm trains over these datasets to predict later at runtime what a user is signing.
8 дек 2014
