Arabic Letter Utterance

Published: 23 April 2021| Version 2 | DOI: 10.17632/r5y668c92d.2
Contributor:
Rian Adam Rajagede

Description

This dataset contains audio recordings taken from 13 men and 13 women. Each subject was asked to read ten different Arabic letters four times. The files are named in the format "XXYYY.wav." - XX is the ID of the read letter (please refer to Guide_Dataset.png) - YYY is the ID of the subject We can see ten letters along with their ID at Guide_Dataset.png. We selected the ten classes based on their similarities: - Class 0 and 1 have similar sounds and properties - Class 2 and 3 have similar sounds and properties - Class 4 and 5 have similar sounds and properties - Classes 6, 7, 8, and 9 have similar sounds and properties

Files

Steps to reproduce

The dataset was taken in different environments but using the same software, Audio Recorder from Sony Mobile Communication. The setting specifications used are the same, mono channel with superior quality (44100Hz sample rate). We reduce the noise using Noise Reduction features of Audacity.

Categories

Applied Sciences

Licence