Bengali Automatic Speech Recognition System

Ahnaf Mozib Samin, M. Humayon Kobir, M. Shahidur Rahman

Dec 2019


ASR is the process of converting an acoustic signal to a sequence of words. In our thesis work, we prepared a 241.2 hours long speech corpus and then we evaluated our corpus with an existing corpus in Bengali Language. We implemented a Convolutional Neural Network based system for the first time in Bengali language. We got 19.86% word error rate with the CNN based model and 23.49% word error rate with the RNN based model.

Type Poster

Location SUST, Bangladesh