A speech dataset of three ethnic languages of Bangladesh: Chakma, Marma and Garo.

This dataset provides a curated collection of speech recordings from three ethnic languages of Bangladesh: Chakma, Marma, and Garo. It contains a total of 2321 WAV audio recordings, each ranging from 1 to 7 seconds in duration. All recordings were collected from 11 native speakers aged 20–26 years, who read from a consistent set of 211 predefined Bengali sentences.

The audio samples were recorded using various smartphone devices across different, naturally occurring acoustic environments, providing diverse background conditions. The dataset includes audio files only, without any accompanying metadata file.

Speaker Distribution

The dataset includes:

Chakma: 5 speakers

Marma: 3 speakers

Garo: 3 speakers

This structured distribution ensures balanced representation across the three ethnic groups.

Key Features of the Dataset

Total Audio Files: 2321

Ethnic Languages: Chakma, Marma, Garo

Total Speakers: 11

Speaker Age Range: 20–26 years

Sentences Read: 211 predefined Bengali sentences

Audio Duration: 1–7 seconds per file

Audio Format: WAV

Recording Devices: Various smartphones

Metadata: No metadata file included; only audio files are provided

Data Organization

All recordings are supplied as individual WAV files. Depending on your folder structure, the files may be grouped by ethnic language or speaker (let me know if you want the folder structure documented in the description).

Potential Applications

This dataset can be used for:

Automatic Speech Recognition (ASR) for low-resource languages

Ethnic language identification (LID)

Speaker identification and verification

Acoustic and phonetic analysis

Training multilingual or cross-lingual speech models

Research on low-resource speech processing

Significance

Ethnic languages in Bangladesh are significantly underrepresented in speech technology research. This dataset offers an important resource for developing speech-based systems and contributing to language preservation, technology advancement, and computational linguistic studies.