Vision-Based Hand Gesture Recognition Using Deep Learning Approach
Hand Gesture Recognition (HGR) serves as a fundamental way of communication and interaction for human being. HGR has the potential to transform Human Computer Interaction (HCI) and also entails many useful applications. For instance, HGR can be used to recognize sign language, which is a visual lang...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Published: |
2019
|
| Subjects: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Hand Gesture Recognition (HGR) serves as a fundamental way of communication and interaction for human being. HGR has the potential to transform Human Computer Interaction (HCI) and also entails many useful applications. For instance, HGR can be used to recognize sign language, which is a visual language and is used all over the world as a primary means of communication by the deaf and mute. Conventional HGR equipment like coloured glove to obtain accurate information about the gesture. It is intrusive, inconvenient and costly, which limits its applicability. Therefore, there is great need for vision-based HGR to simplify the communication and interaction process. However, vision-based HGR faces many challenges such as the variation in background, illumination, hand size, color of skin and also similarities among gestures. Vision-based HGR using traditional machine learning approaches typically involve multiple stages of complicated processing, such as hand-crafted feature extraction methods, which are usually designed to deal with certain challenges specifically. Hence, the effectiveness of the system and its ability to deal with different challenges across multiple datasets are heavily relied on the methods used. In contrast, deep learning approach such as CNN, adapts to different challenges via supervised learning. However, CNN network architecture is not fully explored and exploited for vision-based HGR.
This study investigates the problem of vision-based HGR, with primary focus on deep learning approach. In this thesis, three deep neural network architectures are proposed for vision-based HGR, which are based on variants of Convolutional Neural Network, namely: (1) Wide ResNet, (2) Convolutional Neural Network with Spatial Pyramid Pooling (CNNSPP), and (3) MDenseNet. The proposed methods are evaluated on three public benchmark datasets. |
|---|