Automated Detection and Translation of Multilingual Speech: A System for Real-Time Language Recognition and Conversion

Abstract

Speech is a fundamental aspect of communication, and in India, conversations frequently involve switching between multiple languages. This poses significant challenges for multilingual speech recognition systems, as accurately identifying the language of spoken words, letters, or sentences is complex. The problem is exacerbated by code-switching, where speakers seamlessly alternate between languages. Existing models, often trained on non-representative corpora of Indian languages, struggle with accuracy in these scenarios. To address these challenges, we propose a novel approach that eliminates the need for users to pre-select spoken or transcribed text languages. Our model automatically detects the language in real-time, recognizes the spoken words, and provides output text in all languages used during the conversation. This method simplifies the user experience and improves the accuracy of multilingual speech recognition, making it more effective and user-friendly in multilingual contexts.