Bengali Spoken Digit Classification: A Deep Learning Approach Using Convolutional Neural Network
Riffat Sharmin, Shantanu Kumar Rahut, Mohammad Rezwanul Huq
Abstract
Bengali is a largely spoken language. Bengali speech recognition can have a significant effect in many fields such as human-computer interaction, the internet of things, etc. A part of a Bengali speech recognition system is the process of Bengali spoken digit classification. A few works have been done on Bengali digit classification, but all of them had missed out on one or two influential parameters like dialects, gender or age-groups. Voice of people differs due to gender, dialects, and age. This paper proposes a deep learning approach for classifying the Bengali spoken digits. It takes all parameters like dialects, gender, age-groups into account and the proposed approach acquires more than 98% accuracy using a convolutional neural network (CNN).