Isolated Words Speech Recognition System for Brahvi Language using Recurrent Neural Network

Speech recognition is an interdisciplinary field drawing from computer science, linguistics, and computer engineering. Many modern devices have speech recognition functions in them to allow for easier. The demand for speech recognition in new languages corresponds to cultural preservation, accessibility, market growth, research, enhanced efficiency, innovation, and empowerment, rendering it a significant pursuit within the realm of language technology. Machine learning-based speech-to-text conversion has advanced significantly in the modern era. Brahvi, one of the most spoken languages in Balochistan, is used by approximately 4 million people in Pakistan. In this paper, a cutting-edge model for converting isolated Brahvi words from speech to text using recurrent neural networks (RNN) is proposed. Firstly, we introduced processing techniques and algorithms that can be applied to the speech recognition of single words in the Brahvi language. Seventy regularly used Brahvi language words, encompassing a variety of dialects and voice frequencies, were collected to create a pioneering dataset for this language. The specific RNN model proposed for word speech recognition and its validation on a data set in which the proposed model demonstrates 91 % accuracy.