Bangla Sign in Motion Dataset

Sign language is one of the oldest and most natural forms of language for communication. Individuals who have hearing impairments and/or are nonverbal use sign language as a means of communication. Vision-based sign language recognition is designed to facilitate communication for the deaf, but the available Bangla sign language datasets primarily focus on characters and numbers. This article introduces a valuable resource comprising numerous video recordings of Bangla sign language (BDSL) words, including 133 words performed by 25 signers. Bangla Sign in Motion consists of 33250 videos with 133 labels, each containing 250 videos. Data collection involved capturing videos from 25 different adults, each with unique characteristics and contexts. Various strategies were employed during data preparation to minimise noise. These videos capture individuals using Bengali sign language to communicate through gestures, facial expressions, and body movements. This dataset will be openly accessible to researchers, particularly those involved in vision-based sign language recognition (SLR) and hand gesture recognition (HGR). To the authors knowledge, it is by far the first public BDSL dataset to facilitate word-level sign recognition research.