One-handed Filipino sign language recognition glove prototype using image acquisition and neural networks
The Filipino sign language (FSL) is the primary form of communication between members in the local deaf community. Although FSL is already a growing communication tool used also by hearing individuals, there is still a large number who are unfamiliar with such form of communication. Therefore commun...
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | text |
| Language: | English |
| Published: |
Animo Repository
2008
|
| Subjects: | |
| Online Access: | https://animorepository.dlsu.edu.ph/etd_bachelors/14495 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | De La Salle University |
| Language: | English |
| Summary: | The Filipino sign language (FSL) is the primary form of communication between members in the local deaf community. Although FSL is already a growing communication tool used also by hearing individuals, there is still a large number who are unfamiliar with such form of communication. Therefore communication barriers still exist between deaf and hearing community. The technique for recognizing signs developed in this research allow the creation of a system to serve as a tutorial in learning Filipino sign language.
This thesis details the development of a computer system (labeled the Helping Hand system) capable of recognizing a set of signs from FSL, based on the classification using feed-forward backpropagation neural networks.
The key element in this research is a prototype glove embedded with Flexpoint bend sensors for finger flexion and improvised touch sensors to measure finger contact. By means of image processing through vision system, words are easily recognized with the help of the color coded FSL glove. There are 20 basic FSL words which are present in the Helping Hand Library.
Experiments were conducted to test the accuracy, consistency and performance of the words. Increasing the number of training data for the network used in the recognition process gives better results. With 5 different users, sufficient varying data were obtained. The glove is giving out at least 95.2% accuracy per word with 40 training samples as compare to 84.66% with only 25 samples from all 5 users. All users were able to reach and surpass the 20 word limit for the study.
Keywords: FSL glove, feedforward backpropagation, flexpoint bendsensor, touch sensor, vision system, Helping Hand, sign recognition. |
|---|