Summary
International Conference on Emerging Technologies for Communications
2021
Session Number:D1
Session:
Number:D1-4
Handwritten Numerical Character Recognition using LSTM based on Dual Leap Motion Controllers
Noriaki Kaneko, Masakatsu Ogawa ,
pp.-
Publication Date:2021/12/1
Online ISSN:2188-5079
Summary:
With the increasing infection of the new coronavirus (COVID-19), people expect to introduce the contactless operation rather than the screen touch operation. To address this demand, we focused on the numerical input with contactless. Sine the hand tracking sensor, known as the product name of Leap Motion controller, has the tracking function of the movement of fingers and hands, it is useful for contactless operation. However, it is difficult for most people to communicate the numerical number by their own hands, because they do not know the sign language for hearing-impaired people. Therefore, it is necessary to recognize by the written characters used in daily life. Since the tracking of hand and finger movements differ depending on the location of the Leap Motion controller, we propose to place two Leap Motion controllers at different positions to recognize handwritten characters. We apply the 10 types from 0 to 9 as the handwritten characters and classify these characters by machine learning algorithms such as LSTM. As a result of the classification accuracy, applying the bidirectional LSTM classification achieves the maximum, 94.7%.