Telugu Spoken Digits Modeling using Convolutional Neural Networks

Abstract

Automatic speech recognition is an attractive problem currently because of its applications in various areas such as IoT, Interactive voice recognition systems, mobile interfaces, etc. Digit recognition problem problem is an important task for aforesaid applications. Since Telugu is a low-resourced language, the recognition problem has been addressed in a limited way. The spoken digit recognition is handled using Convolutional Neural Networks (CNNs) by understanding the nature of speech signals through Mel Frequency Cepstral Coefficients (MFCCs). We observed the approach performed with a recognition rate of 73%. The proposed approach and the obtained results are discussed in the paper.