KNN-based Speech-to-Text Conversion for Bangla to Enhance Regional Language Processing

In order to close the digital divide and promote greater linguistic diversity in natural language processing applications, speech to text conversion for regional languages is a crucial first step. This study introduces a novel method for employing the K-Nearest Neighbors (KNN) algorithm to translate spoken Bangla, a regional language that is mostly spoken in Bangladesh and West Bengal, India, into text. By mapping audio data to linguistic representations, the suggested method uses the KNN's pattern recognition ability to convert Bangla voice into text. To extract the discriminative aspects of spoken language, we use the Mel-frequency cepstral coefficients (MFCCs), which are crucial audio features. To improve the model's resilience, our dataset includes a wide range of Bangla speech samples, including regional dialects and accents. This work contributes to the broader goal of democratizing speech technology for regional languages, empowering non-English-speaking communities, and making information accessible to a wider audience. The proposed KNN-based system opens the door to diverse applications of regional language interfaces.