Extreme learning machine based general purpose real time hand gesture recognition
With the technology advancement of 3D cameras and computational power, hand gesture based Human Machine Interaction (HMI) is recognized to be the future trend for Natural User Interface (NUI). However, one of the bottlenecks for current development of hand gesture recognition is the accuracy and rea...
Saved in:
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Final Year Project |
| Language: | English |
| Published: |
2014
|
| Subjects: | |
| Online Access: | http://hdl.handle.net/10356/61436 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Nanyang Technological University |
| Language: | English |
| Summary: | With the technology advancement of 3D cameras and computational power, hand gesture based Human Machine Interaction (HMI) is recognized to be the future trend for Natural User Interface (NUI). However, one of the bottlenecks for current development of hand gesture recognition is the accuracy and real-time performance (prediction speed) of classification algorithms. To tackle this challenging problem, the FYP thesis “Extreme Learning Machine (ELM) based general purpose real-time hand gesture recognition” is proposed. In this report, the following research and developments are discussed: feature extraction using Relative Position and Rotation Method (RPRM), Time Window Method (TWM) proposed by the author, ELM performance evaluation (accuracy and speed), real-time static gesture recognition application, real-time dynamic gesture recognition application, real-time motion recognition application and real-time hand-gesture desktop control application. RPRM and TWM are proven to be effective feature extraction method and ELM performs accurately and speedy in real-time using the C++ coded applications listed above. This project is part of the NTU-BMW joint research project led by Professor Huang Guangbin. |
|---|