Creating a Comprehensive Dataset for Older Bengali Speakers’ Voices: Towards Enhanced Speech Recognition and Assistive Technologies

The aging population presents unique challenges for speech recognition technologies, particularly for languages like Bengali, where limited resources exist for elderly speakers. This paper addresses the gap in Bengali speech recognition for older adults by creating a comprehensive dataset that captures the vocal characteristics of elderly Bengali speakers. The dataset includes diverse speakers with variations in age, gender, dialect, and speech tasks, reflecting the full spectrum of age-related speech variations. The methodology involves collecting and pre-processing audio from publicly available resources, including YouTube, supplemented by contributions from Bangladesh Betar, followed by noise reduction and audio segmentation. This dataset aims to enhance the accuracy of speech recognition systems and assistive technologies tailored to elderly Bengali speakers, fostering inclusivity and improving accessibility. The findings suggest that incorporating diverse elderly speech data is critical for developing more effective speech technologies for this demographic. This work contributes to bridging the gap in speech technology for older Bengali speakers and paves the way for future research in the field of assistive technologies and speech recognition.