A Survey of the Effects of Data Augmentation for Automatic Speech Recognition Systems

Abstract

Data augmentation has been proposed as a method to increase the quantity of training data. It is a common strategy adopted to avoid over-fitting, reduce mismatch and improve robustness of the models. But, would the system performance improve if we add data of any nature? This paper presents a survey about data augmentation techniques and its effect on Automatic Speech Recognition systems, some experiments were carried out to support the hypothesis that adding noise is not allways help.