TELUGU ANKELU: A Telugu Spoken Digits Corpora for Mobile Speech Recognition

Automatic Speech Recognition (ASR) deals with recognizing spoken words or sentences through an automated program or an algorithm. ASR has multi-faceted applications in various domains such as Mobile Speech Recognition, IoT, Human-Machine interaction, etc. Researchers have been working on different problems in these domains for more than 5 decades. Mobile speech recognition is a pre-dominant area that supports a wide variety of applications useful for physically challenged, senior citizens and novice mobile users. Though the domain has significant practicality, it faces challenges due to the lack of data in the targeted language. Though it is not major for well-developed languages like English, it is very profound in low-resourced and under-resourced languages. India has 1,391.99 million inhabitants speaking a wide variety of languages with Hindi being the most popular language. There are many native languages spoken in each region of the country. But, research that contributes towards the development of ASR is questionable due to many reasons, and the unavailability of data is one of those. Hence, an open-source speech digit data set has been prepared and will be available through this paper for the Telugu language which is the sixth largely spoken language in the country.