Scholarlink

<p>Sign language discourse is an essential mode of daily communication for the deaf and hard-of-hearing people, yet, research on Bangla Sign Language (BdSL) faces significant limitations, primarily due to the scarcity of datasets. Recognizing word-level signs in BdSL (WL-BdSL) presents multifaceted challenges like proper annotation procedure, capturing the dynamic nature of sign gestures from facial or hand landmarks, developing suitable machine learning or deep learning-based models with substantial video samples, and so on. In this paper, we aim to address these challenges by introducing BdSLW60, a comprehensive BdSL word-level dataset captured in an unconstrained and natural setting (allowing positional, temporal variations, and changes in hand dominance). BdSLW60 comprises 60 Bangla sign words, with a significant scale of 9307 video trials provided by 18 signers under the supervision of a sign language professional. The dataset was rigorously annotated and cross-checked by 60 annotators, ensures the dataset’s quality and reliability. Additionally, we propose a novel relative quantization-based key frame encoding technique for landmark-based sign gesture recognition. We report the benchmarking of our BdSLW60 dataset using the Support Vector Machine (SVM) with testing accuracy up to 67.6% and an attention-based bi-LSTM with testing accuracy up to 75.1%. The availability of BdSLW60 dataset (<a href="https://www.kaggle.com/datasets/hasaniut/bdslw60">https://www.kaggle.com/datasets/hasaniut/bdslw60</a>), along with the associated code base (<a href="https://github.com/hasanssl/BdSLW60_Code">https://github.com/hasanssl/BdSLW60_Code</a>), promises to enhance further research and applications in the realm of Bangla Sign Language.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

BdSLW60: A word-level bangla sign language dataset

  • 期刊论文

  • 发表日期：

    2025年4月14日

查看全文
  • Husne Ara Rubaiyeat,
  • Hasan Mahmud,
  • Ahsan Habib,
  • Md Kamrul Hasan

摘要

Sign language discourse is an essential mode of daily communication for the deaf and hard-of-hearing people, yet, research on Bangla Sign Language (BdSL) faces significant limitations, primarily due to the scarcity of datasets. Recognizing word-level signs in BdSL (WL-BdSL) presents multifaceted challenges like proper annotation procedure, capturing the dynamic nature of sign gestures from facial or hand landmarks, developing suitable machine learning or deep learning-based models with substantial video samples, and so on. In this paper, we aim to address these challenges by introducing BdSLW60, a comprehensive BdSL word-level dataset captured in an unconstrained and natural setting (allowing positional, temporal variations, and changes in hand dominance). BdSLW60 comprises 60 Bangla sign words, with a significant scale of 9307 video trials provided by 18 signers under the supervision of a sign language professional. The dataset was rigorously annotated and cross-checked by 60 annotators, ensures the dataset’s quality and reliability. Additionally, we propose a novel relative quantization-based key frame encoding technique for landmark-based sign gesture recognition. We report the benchmarking of our BdSLW60 dataset using the Support Vector Machine (SVM) with testing accuracy up to 67.6% and an attention-based bi-LSTM with testing accuracy up to 75.1%. The availability of BdSLW60 dataset (https://www.kaggle.com/datasets/hasaniut/bdslw60), along with the associated code base (https://github.com/hasanssl/BdSLW60_Code), promises to enhance further research and applications in the realm of Bangla Sign Language.