Text Generation through Hand Gesture Recognition
14 Pages Posted: 23 Jun 2020
Date Written: June 23, 2020
Abstract
Development of Vision-based Hand Gesture Recognition systems has proved to be beneficial for disabled people over the past. These systems are user friendly and inexpensive. They also make room for contemporary endeavours in the Human-Computer Interaction area. Gestures can control the cursor, the music player, or even play games. Despite all the improvements made in Vision-based recognition systems, a fast, accurate, and reliable system is yet to see the light of the day. This study aims to understand the previous work in this field and develop a fast and accurate alternative to perform the same task. To render the system faster, we propose to replace the histograms from the previous approach and uses native OpenCV functions. The Use of these functions can reduce the time for calculating the histogram and loading it when required. The final system developed, succeeded in predicting 37 gestures with a 98% accuracy. The system also has a calculator mode, where some gestures are reserved for operators and others for operands. This study tells us how we can use different computer vision techniques to make the recognition process more memory efficient and lightweight. In the future, this could be used as a commercially viable option by putting it on app stores.
Keywords: deep learning, computer vision, hand gesture recognition, convolutional neural networks, thresholding, morphological closing
Suggested Citation: Suggested Citation