| Abstract: | We are developing a lightweight wearable American Sign Language translation device consisting of 5 wireless fingernail devices, 5 ring devices, and a wrist device. The ring devices are composed of three infrared (IR) emitter/receiver pairs that are operated by a microcontroller in the wrist device. All of the IR LEDs flash in succession, which will trigger the IR LEDs in the wireless fingernail devices to flash. The intensity of the signal emitted by the LEDs in the ring and fingernail devices is collected by the sensors. A light intensity matrix is then created to determine the position of each finger. The microcomputer in the wrist device uses this matrix to train the neural network developed to predict the gesture. Lastly, an audio output of the translated sign is initiated allowing anyone near the user to understand the gesture. Four buttons will allow the user to reset to a default training, retrain the neural network, translate their sign, or restart the system. The device works without any external monitor or computer making it unencumbering and easy to use. |