Related Experiment Video
Updated: Jul 8, 2025
05:41
A Step-by-Step Implementation of DeepBehavior, Deep Learning Toolbox for Automated Behavior Analysis
Published on: February 6, 2020
9.4K
BdSL47: A complete depth-based Bangla sign alphabet and digit dataset
S M Rayeed1, Sidratul Tamzida Tuba1, Hasan Mahmud1
1Systems and Software Lab (SSL), Department of Computer Science and Engineering (CSE), Islamic University of Technology (IUT), Board Bazar, Gazipur 1704, Bangladesh.
Data in Brief
|December 11, 2023
Summary
This study introduces BdSL47, a new depth dataset for Bangla Sign Language (BdSL) to improve communication for the deaf-mute community. An Artificial Neural Network (ANN) model achieved a 97.84% F1 score, demonstrating effective sign recognition.
Area of Science:
- Computer Science
- Artificial Intelligence
- Human-Computer Interaction
Background:
- Sign Language Recognition (SLR) is vital for communication accessibility.
- Developing comprehensive sign language datasets, especially for Bangla Sign Language (BdSL), is challenging due to gesture complexity and limited depth data.
- Accurate BdSL recognition is hindered by the scarcity of specialized datasets.
Purpose of the Study:
- To introduce BdSL47, an open-access depth dataset for 47 static one-handed BdSL signs (10 digits, 37 letters).
- To develop and evaluate an Artificial Neural Network (ANN) model for BdSL recognition using depth data.
- To facilitate advancements in BdSL recognition through a publicly available, comprehensive dataset.
Main Methods:
- Created the BdSL47 dataset using the MediaPipe framework to extract depth information from 47 static BdSL signs.
- Developed an Artificial Neural Network (ANN) model with specific architecture (63-node input, 47-node output, 4 hidden layers with dropout, Adam optimizer, ReLU activation).
- Trained and evaluated the ANN model on the BdSL47 dataset to classify gestural inputs.
Main Results:
- The proposed ANN model achieved a high F1 score of 97.84% on the BdSL47 dataset.
- The model demonstrated effective learning of spatial relationships and patterns from depth-based gestural features.
- The results indicate superior performance compared to existing baseline methods.
Conclusions:
- The BdSL47 dataset provides a valuable resource for advancing Bangla Sign Language Recognition.
- The developed ANN model shows significant effectiveness in recognizing BdSL signs from depth data.
- The availability of this dataset is expected to spur further research and development in deep learning for BdSL.
