An Isolated Words Balanced Corpus for Native and Non-Native Urdu Speakers in Automatic Speech Recognition
Shalini V. Sathe, Ratnadeep R. Deshmukh, Santosh K. Maher, Swapnil D. Waghmare · 2023
This study explores the fusion of speech technology and innovation, enriching human language nuances for connectivity and barrier-free communication. Committed to collective progress, we aim to unlock potential through an Urdu voice corpus. This dataset, focusing on spontaneous speech, includes seventeen isolated Urdu words spanning numeric digits and days of the week. Utilizing advanced tools like microphones and PRAAT software, diverse speaker recordings capture vocal intricacies. With participants aged 20 to 40, we create a reliable dataset in a controlled, noise-free environment. Embracing both native and non-native Urdu speakers, our work addresses the underrepresentation of Urdu in isolated word recognition. This research amplifies Urdu’s presence in speech technology and advocates for linguistic inclusivity, fostering a harmonious symphony of diversity.