Recognition of Different Voice Pathologies with Data Augmentation Strategies

Abstract

                In the past few years, there is a lot of improvement in performance of speech analysis methods. Especially, deep learning algorithms plays a crucial role in speech synthesis. Different voices of different genders are going to be classified with the help of deep CNN methodology are mainly focused in this research work. For identifying voice pathologies, different deep learning algorithms are most suitable and for performing this task, there is a lot of requirement of training datasets. But there are limited resources of real-world audio data and here the different audio augmentation methods are to be used for increasing the training dataset. Speech augmentation is mainly used for the training of ANN and it will make the predictions effective. This Strategy is also going to avoid the overfitting and it will improve the model’s robustness. Speech classification has shown a improved accuracy when compared to the existing models.