Fighting COVID-19: Convolutional Neural Network for Elevator User’s Speech Classification in Bahasa Indonesia

Abstract

The use of deep learning to automatically recognize and classify the floor numbers and commands spoken by elevator users could help reduce transmission of COVID-19 by physical contact with the elevator button. Fortunately, the ability of Convolutional Neural Network (CNN) as one of the deep learning architectures to recognize patterns is well-known. This research aims to create isolated word data of spoken floor numbers and commands, then build a classifier model to recognize and classify floor numbers and commands spoken by elevator users. In this research, speech data are gathered in Bahasa Indonesia and classified using CNN and Multilayer Perceptron (MLP). At the end of this research, it is found that 94% of classification accuracy is provided by the best CNN model configuration towards test data. This outcome is better than the MLP model which provides 80% of accuracy.