Robust Speech Recognition System in Noisy Environment Using Mel Frequency Cepstrum Coefficient and Dual Local Binary Pattern

  • Bhuvaneshwari Jolad, Dr. Rajashri Khanai

Abstract

Speech recognition is crucial part of many modern artificial intelligent based automation systems. Most of the speech recognition systems are facing the challenges of higher word error rate because of noisy database, poor feature discrimination and weak classifiers. In this work, robust speech recognition system is proposed using Mel Frequency Cepstrum Coefficient (MFCC) and Double Local Binary Pattern (DLBP) is used for handling the temporal variations in the speech signal. Neural network classifier is used for the classification. Extensive experiments are carried out on the English vocabulary database of 10 words and performance is evaluated on the basis of percentage recognition accuracy. Four types of real time noises such as street, train, party crowd and wind noises are used for the noisy training and It is observed that the proposed scheme gives promising results in case of noisy training with the recognition accuracy of 83.88 % and for clean training it gives 93 % average recognition accuracy.

Published
2020-06-06
How to Cite
Bhuvaneshwari Jolad, Dr. Rajashri Khanai. (2020). Robust Speech Recognition System in Noisy Environment Using Mel Frequency Cepstrum Coefficient and Dual Local Binary Pattern. International Journal of Advanced Science and Technology, 29(04), 7498 -. Retrieved from http://sersc.org/journals/index.php/IJAST/article/view/28161