Spoken Digit Classification using Deep Learning Algorithms

Abstract

The deep learning technique uses speech recognition in many different applications, including voice assistants, voice authentication, audio transcriptions, etc. Children who are dyslexic, blind persons and those with impairments can all benefit from spoken digit recognition. The goal of this paper is to create spoken digit recognition for the categorization of digits from 0 to 9 utilizing Convolution Neural Networks (CNN) and Long Short -Term Memory neural networks. With the addition of autoencoders, the performance of the CNN model is assessed. Finally, a comparative analysis is performed on the performances of the models based on the performance metrics.