Applications of Deep Learning Approaches in Speech Recognition: A Survey

Abstract

Automated speech recognition (ASR) appeared to be a driving force for variety of machine learning (ML) techniques, include to ubiquitously utilized discriminative learning, Bayesian learning, hidden Markov model, adaptive learning, and structured sequence learning. Although machine learning utilize ASR as a large scale, it can reasonable application to thoroughly test viability for a given procedure and to motivate unused issues emerging from intrinsically consecutive and discourse energetic nature. Also, although ASR is accessible commercially for a few applications used in this research through the limitation and research gaps that the researcher try to access high accuracy of these systems. The advance technology from new ML techniques appears incredible guarantee to progress the literature review in ASR innovation. This study gives reader with a diagram of present-day ML methods as used within the relevant and current as significant for ASR future systems and research. The study goal is to promote advanced cross-pollination between ML and ASR communities more than has hither to occurred.