Recognition of Holoscopic 3D Video Hand Gesture Using Convolutional Neural Networks

Norah Alnaim,Rafiq Swash,Maysam Abbod

doi:10.3390/technologies8020019

Norah Alnaim, Rafiq Swash + Show 1 more

Open Access

https://doi.org/10.3390/technologies8020019

Copy DOI

Journal: Technologies	Publication Date: Apr 15, 2020
Citations: 2	License type: CC BY 4.0

Affiliation: Brunel University London

Abstract

The convolutional neural network (CNN) algorithm is one of the efficient techniques to recognize hand gestures. In human–computer interaction, a human gesture is a non-verbal communication mode, as users communicate with a computer via input devices. In this article, 3D micro hand gesture recognition disparity experiments are proposed using CNN. This study includes twelve 3D micro hand motions recorded for three different subjects. The system is validated by an experiment that is implemented on twenty different subjects of different ages. The results are analysed and evaluated based on execution time, training, testing, sensitivity, specificity, positive and negative predictive value, and likelihood. The CNN training results show an accuracy as high as 100%, which present superior performance in all factors. On the other hand, the validation results average about 99% accuracy. The CNN algorithm has proven to be the most accurate classification tool for micro gesture recognition.

Highlights

A major form of interaction between users and computers is achieved through devices like the mouse, keyboard, touchscreen, remote control, and other direct contact methods.Communication amongst humans is achieved through more intuitive and natural non-contact methods, e.g., physical movements and sound
Twelve different 3D micro hand gestures for three subjects are fed as input into convolutional neural network (CNN)
The comparison, including execution time, accuracy, specificity, sensitivity, positive predictive value PPV, negative predictive value (NPV), likelihood, and RMS, is represented in specificity, sensitivity, positive predictive value PPV, NPV, likelihood, and RMS, is represented in noted that, for WT, the execution time is less than the total time execution of CNN and empirical mode decomposition (EMD)

Summary

Introduction

A major form of interaction between users and computers is achieved through devices like the mouse, keyboard, touchscreen, remote control, and other direct contact methods. Communication amongst humans is achieved through more intuitive and natural non-contact methods, e.g., physical movements and sound. The efficiency and flexibility of these non-contact interaction methods have led several researchers to consider using them to support human–computer communication. Gesture forms a substantial part of the human language. It is an important non-contact human interaction method. The primary goal of such gesture recognition is to develop a system that can recognize and understand specific gestures and communicate information without any human intervention. The use of hand gestures for a human computer interface (HCI) offers direct measurable inputs by the computer [1]. Using a controlled background makes hand gesture detection easier [2]

Objectives

Methods

Results

Conclusion