Speech To Text Using Deep Learning

Abstract

The surge in huge audio data and progress in deep learning technologies has opened up possibilities for the development of highly accurate Automatic Speech Recognition (ASR) systems. These systems promise to automate transcription, provide real-time accessibility, and ultimately empower seamless Human-Computer Interaction (HCI). This paper outlines a streamlined workflow for training and deploying neural ASR models. We synthesize findings from previous studies on statistical modelling in the auditory domain, detailing of our system's architecture using Composite Neural Framework, and report on experimental results from simulation studies. Our findings suggest that integrating advanced audio processing with intelligent models which helps to reduces the Word Error Rate (WER) and enhances accessibility.